In This Issue
Fall Bridge Issue on Engineering, Technology, and the Future of Work
September 15, 2015 Volume 45 Issue 3

Autonomous Vehicles Implications for Employment Demand

Tuesday, September 15, 2015

Author: Jennifer M. Miller

Autonomous vehicle (AV) technology, popularly envisioned as “driverless cars,” has reached maturity and the cusp of commercialization (RAND 2014; Urmson 2015). Attention has turned to the many policy issues raised by such vehicles on public roads (Beiker 2012; Khan et al. 2012). Reports have largely focused on the use of autonomous vehicles for commuting; less attention has been paid to how they will be used on the job. Yet businesses may be early adopters of AV technology, as an extension of existing forms of workplace automation.

The business case for adoption of AV technology is strong. Most of the more than 35,000 annual traffic deaths in the United States involve driver error (RAND 2014), and the National Highway Traffic Safety Administration (NHTSA 2003) estimates that on-the-job motor vehicle accidents cost US employers almost $55 billion each year.1 Furthermore, as explained below, in addition to saving lives and reducing insurance costs, lost work time, and property damage, autonomous vehicles will save labor. Industry, affected workers, policymakers, and those preparing the future workforce should therefore consider the potential impacts of autonomous vehicles on employment. Data from the Department of Labor (DOL) and the Bureau of Labor Statistics (BLS) indicate which occupations are likely to face the biggest changes.

Economic Perspectives on Automation

Economists view automation as a substitution of capital for labor in a production process. When capital becomes less expensive relative to labor, production processes tend to rely more on automation and less on human labor. The level of substitution varies by industry (Arrow et al. 1961).

When labor and capital are substitutes, if capital becomes less expensive, fewer workers will be employed and they will be paid less. For example, if a road improvement reduces travel time between two cities, drivers will be employed for fewer hours to conduct the same level of trade. Even if trade increases, the reduction in hours due to improved conditions may still outweigh the benefits of the increased trade.

More recently, economists have looked at how technology complements, rather than substitutes for, certain types of high-skilled labor (Acemoglu and Autor 2011). For example, by increasing workers’ productivity, personal computers increased employment and earnings of workers with complementary skills. Consider the effect of computers on the employment of quality control analysts: as computers facilitate measurement, data analysis, and communication, workers with complementary skills may be hired and better compensated to do quality control work in more settings.

Potential Employment Effects of Autonomous Vehicles

To understand how AV technology will affect employment, it is helpful to look at the tasks and work activities associated with specific occupations2 (Autor et al. 2003).

Driving has typically been seen as a manual, nonroutine task and therefore not very amenable to automation. With the maturity of AV technology, it may be time to revisit that assumption. NHTSA describes vehicle automation along a continuum ranging from level 0—all functions are controlled by a human driver—to level 4—the vehicle can drive itself. Employment may begin to be significantly affected at level 3, where automation allows the driver to safely do other work.

With widespread adoption of AV technology, many vehicle trips may no longer require a human driver. And for those that do, the work activities and required skills and abilities will change.

Occupational Data on Driving as a Work Activity

The Occupational Information Network (O*NET; www.onetonline.org) is the US government’s official source of information about occupations, with data from surveys of workers, occupational experts, and occupational analysts. Collected under the sponsorship of the DOL and based on the Standard Occupational Classification, the data include detailed descriptions of the tasks, work activities, and tools (a category that includes vehicles) used in each occupation. It is thus possible to focus on occupations that use relevant types of vehicles and examine possible employment impacts from the adoption of AV technology.

Several elements of the O*NET database may be useful for identifying occupations that are likely to be affected by AV technology.3 Of particular interest, the description of each occupation’s work context includes the extent to which work takes place “in an enclosed vehicle or equipment.” Occupations are scored from 0 to 100 based on responses on a scale from “never” to “every day.”

Scores are similarly assigned for the importance and level of activity involved in “Operating Vehicles, Mechanized Devices, or Equipment” based on a scale that assigns points for “operate a car” (approximately 30 points), “drive an 18-wheel tractor-trailer” (approximately 60), and “hover a helicopter in a strong wind” (approximately 85). This analysis focuses on 85 occupations that score above 50 on both the operating vehicles importance scale and the enclosed vehicle work context scale.

The potential for substitution and complementarity between capital-intensive AV technology and the labor of human drivers raises a host of questions about impacts on the workforce. Which workers are likely to lose their jobs? For which jobs will driving no longer be a required skill, increasing the labor pool from which these workers could be recruited? Which complementary skills will be in greater demand if autonomous vehicles become widely adopted? Which occupations and regions are most likely to be affected?

  Table 1 

To frame discussion of these questions, this analysis classifies occupations into four types: drivers, visitors, hosts, and teams (table 1). The typology differentiates between occupations in which driving is the primary work activity and those in which it is an important but secondary work activity. In either case, the autonomous vehicle can be a substitute for or a complement to the worker’s labor.

Classification of Occupations

Occupations likely to be significantly affected by adoption of AV technology are those that place a high importance on operating vehicles and those that involve working regularly in an enclosed vehicle. O*NET data make it possible to illustrate the categories of occupational vehicle use, as shown in table 2.

  Table 2 

The four types—driver, visitor, host, and team—are described more thoroughly in the following sections, which include the criteria used to define each type; a representative example occupation; likely labor market effects, including competition and skill requirements; and expected technical challenges and opportunities for incorporating autonomous vehicles into the type of occupation.

Drivers

For this category of occupation, driving is the primary task, typically done under routine conditions and with few other tasks, none of which require a high level of skill. A representative job in this category would be pizza delivery driver; others might be heavy and tractor-trailer truck drivers, couriers and messengers, and refuse and recycling collectors.

In these jobs there is little opportunity for the human driver to add value either in transit or at the worksite, so the jobs might be fully automated, removing the human driver from the occupation entirely. For example, Amazon has filed a patent application (number 14/502707) for an unmanned aerial vehicle delivery system, informally referred to as an “Amazon drone.”

If driving becomes a fully automated activity, capital could substitute for labor almost completely in these occupations. Technical aspects of incorporating autonomous vehicles would focus on automating all additional tasks, such as identity verification in the case of a delivery worker, that are currently performed by the human driver. Without a human driver, these vehicles could be optimized for cargo space, fuel efficiency, and other criteria.

Visitors

Visitor occupations may involve a lot of time driving and place a high importance on the ability to drive, but the driving is routine and the worker’s productive activity takes place at a worksite. Representative occupations in this category are pest control workers, heating and air conditioning mechanics and installers, and tree trimmers or pruners.

The visitor’s opportunity to be productive while in transit is minimal. Assuming the work performed onsite is not routine enough to be automated, visitor jobs would still require a human worker to travel from site to site, but that worker would no longer need to be willing—or perhaps even able—to drive a vehicle as a significant work activity. These occupations would then be open to workers now excluded because of the driving requirement, whether for legal, health, or personal preference reasons.

For onsite work that requires a relatively low skill level, current workers may have been selected in large part for their driving ability. They should expect to face more labor market competition if autonomous vehicles are widely adopted.

The technical aspects of adapting autonomous vehicles for these occupations may depend on whether the workers are self-employed. If they are, these aspects may involve entertainment amenities for enjoyment between jobs. For firms that employ workers in visitor occupations, technical aspects are likely to balance risk management by monitoring employee activity in the vehicle (e.g., the presence of unauthorized passengers, substance use) with amenities (e.g., entertainment systems, food service) that may be incorporated to cost-effectively improve employee retention and engagement.

Hosts

In host occupations, driving is the primary work activity. Some of it may be nonroutine. Representative Figurein this category are school bus drivers and other bus or shuttle drivers for schools, airports, and special clients, such as tour guides.

Workers in these jobs are typically selected based on driving skill and their attention is fully occupied with driving. With an autonomous vehicle, there would be an opportunity for the worker to add value in transit. A program in Hartsville, South Carolina, suggests an intriguing possibility (Chaltain 2015): As part of a school development program, school bus drivers were trained in the basics of child development. The program led to a 71 percent reduction in disciplinary referrals and other positive outcomes for students, families, and teachers. This program clearly demonstrates that if the bus is largely autonomous, school bus drivers can be trained or selected for their ability to work with children, turning travel time into learning time.

Workers in host occupations should be prepared for changing skill demands. The technical aspects of adapting autonomous vehicles for these occupations would likely involve the addition of features related to skills that the affected workers do not currently use, providing some opportunity for creativity and innovation but also introducing the potential need to develop new skills.

Teams

In the team category, the worker and vehicle have complementary “skills and abilities” to achieve high productivity. Driving is an important skill but not necessarily an everyday work activity. Workers are typically selected based on a skill other than driving, although lack of ability or willingness to drive is likely to exclude someone from the job. If automation frees these workers from driving, they will be able to use their current skills productively while in transit.

Representative occupations in this category are ambulance drivers, emergency medical technicians (EMTs), and paramedics. The category might also include criminal investigators and special agents, real estate sales agents, and park naturalists.

With sufficient automation of the transportation infrastructure, including networked communication to traffic signals and other vehicles, emergency response vehicles could be highly autonomous. EMTs and paramedics would be free to prepare for the emergency situation on the way to the site and care for patients while transporting them to the hospital. Workers in these occupations should prepare for more intense labor market competition, increasing skill requirements, and potentially greater labor market demand due to enhanced productivity. For example, the ability to fully staff an ambulance with one person could enable greater coverage in rural areas or during off-hours.

The technical aspects of introducing autonomous vehicles in team occupations would focus on maximizing worker productivity while in transit. Adaptations could include the incorporation of considerable additional technology into the vehicle, to the point that the vehicle functions effectively as an office, medical facility, or lab.

Size and Distribution of Employment Effects of Autonomous Vehicles

Data from the BLS Occupational Employment -Statistics (OES) make it possible to quantify the number of US workers and identify the states most likely to be -affected by autonomous vehicle use.

Based on May 2014 estimates, approximately 11.3 million workers, representing 9 percent of total employment, are in one of the 85 occupations likely to be significantly affected by autonomous vehicles.4 Ten occupations account for almost 60 percent of those 11.3 million employees (table 3).

  Table 3 

Variations in regional economies will also determine the effects of autonomous vehicles on employment. Figure 1 shows that potentially affected workers are most concentrated in states with low population density where extractive industries make up a significant part of the economy, based on a location quotient (LQ) that measures such concentration.5 A higher LQ indicates a higher share of potentially affected workers than the state’s total labor force. North Dakota, Wyoming, and West Virginia are the states with the highest LQs.

  Figure 1 

A breakdown of occupations in the three most affected states shows that workers in the driver and visitor categories will bear the brunt of the impacts of AV technology (figure 2). In contrast, California has the lowest LQ, with a smaller share of potentially affected workers in the extractive sector and comparatively more in law enforcement and light truck and delivery services.

  Figure 2 

California’s low LQ may at first seem surprising, since the state has been at the forefront of developing and testing autonomous vehicles. But analysis of the data suggests that, although the commuting benefits of -autonomous vehicles are attracting attention in big -cities on the coasts, the occupational impact may be greater in the Mid- and Mountain West: although individuals are highly motivated to avoid complex and congested urban driving, rural driving may be easier to automate.

Additional Considerations

There are three main considerations to keep in mind to put this very preliminary, even speculative, analysis in context. First, it is specific to the United States. The impact of autonomous vehicles will be global, and other countries, such as Singapore, may lead in adoption (Land Transport Authority 2015). The extent to which capital substitutes for labor will vary with the makeup of a national economy (Arrow et al. 1961), and this will likely apply to the adoption of autonomous vehicles as well.

Second, it is difficult to disentangle the potential effects of autonomous vehicles from other types of industrial automation. This difficulty manifests in two ways.

In a practical sense, with existing occupational data it is not always possible to distinguish the operation of road-based delivery or transport vehicles versus agricultural or industrial equipment. Although the O*NET data on “tools” used in each occupation include type of vehicle used (e.g., cars, light trucks or sport-utility vehicles, and minivans or vans), they seemed incomplete and so were used for confirmation and clarification only.6 It would be useful for future occupational surveys to ask specific questions about driving as a work activity and vehicles as tools and technology. This information would be useful to understand how occupations may change at this new frontier of workplace automation.

More broadly, it is possible that other forms of automation will render autonomous vehicles irrelevant. For example, widespread adoption of e-learning and telecommuting might drastically reduce the need to transport large groups of children to school on a regular basis. The school bus driver may lose out to e-learning rather than to the self-driving bus.

Third, this analysis cannot account for new occupations that may emerge—for example, in the building, management, and maintenance of autonomous vehicle systems. But it also suggests some ways in which a wide variety of jobs might be transformed with the adoption of autonomous vehicles: Demand for certain transportation and delivery work may be drastically reduced with the advent of fully autonomous vehicles. Jobs that involve skilled or manual labor performed at multiple or remote worksites may no longer involve driving as a significant work activity, thus expanding the labor supply. These effects may be particularly strong in rural states with extractive economies.

Finally, some skilled workers may find that their autonomous vehicle includes occupation-specific technology to enable them to maximize productivity while in transit. Other workers may find that they must develop additional complementary skills to remain competitive once driving becomes a less marketable skill.

Conclusion

There are two important reasons to continue examining the occupational impacts of AV technology. Knowledge of these effects can foster innovation by revealing industrial and geographic market opportunities. The earliest adopters of mobile phones, for example, were people with specialized occupational needs. AV technology may follow a similar path.

Better information about the occupational effects of this technology is also important for workforce planning. As manufacturing became increasingly automated and moved offshore, many US workers and communities lost their productive roles in the economy and are still struggling to recover. With improved information and foresight, governments, educational institutions, and individuals can make better labor market decisions as autonomous vehicles improve safety and autonomy both on the road and on the job.

References

Acemoglu D, Autor D. 2011. Skills, tasks and technologies: Implications for employment and earnings. Handbook of Labor Economics 4:1043–1171.

Arrow KJ, Chenery HB, Minhas BS, Solow RM. 1961. Capital-labor substitution and economic efficiency. Review of Economics and Statistics 43(3):225–250.

Autor DH, Levy F, Murnane RJ. 2003. The skill content of recent technological change: An empirical exploration. Quarterly Journal of Eco-nomics 118(4):1279–1333.

Beiker SA. 2012. Legal aspects of autonomous driving. Santa Clara Law Review 52(4):1145–1156.

Chaltain S. 2015. A town where a bus is more than a bus. New York Times, February 27.

Khan AM, Bacchus A, Erwin S. 2012. Policy challenges of increasing automation in driving. IATSS Research 35(2):79–89.

Land Transport Authority. 2015. Singapore Autonomous Vehicle Initiative. News release, January 13. Available at www.lta.gov.sg/apps/news/page.aspx?c=2&id=453eb7c1- bb81-4243-a77a-d30d2c811b9b.

NHTSA [National Highway Transportation and Safety Administration]. 2003. The Economic Burden of Traffic Crashes on Employers: Costs by State and Industry and by Alcohol and Restraint Use. Washington.

RAND. 2014. Autonomous Vehicle Technology: A Guide for Policymakers. Santa Monica: RAND Corporation Transportation, Space, and Technology Program.

Urmson C. 2015. Progress in self-driving vehicles. The Bridge 45(1):5–8. 

FOOTNOTES

1 In 2015 dollars, based on data from 1998–2000.

2 Tasks and work activities are two elements of the six-part Occupational Information Network (O*NET) content model used to describe occupations. Tasks are occupation-specific and work activities cross occupations.

3 The O*NET scales include watercraft, airborne vehicles, and manufacturing equipment; these are not relevant to this analysis, which focuses on jobs that use vehicles suitable for travel on public roadways. All discussion and analysis in this article are specific to occupations likely to operate vehicles on public roads.

4 Data from the May 2014 BLS OES estimates (available at www.bls.gov/oes/). Self-employed workers and four occupations with no clear mapping from O*NET to the OES are excluded from this total.

5 The Bureau of Economic Analysis (www.bea.gov/faq/index.cfm?faq_id=478) provides the following explanation of a location quotient: “A location quotient (LQ) is an analytical statistic that measures a region’s industrial specialization relative to a larger geographic unit (usually the nation). [It] is computed as an industry’s share of a regional total for some economic statistic (earnings, GDP by metropolitan area, employment, etc.) divided by the industry’s share of the national total for the same statistic. For example, an LQ of 1.0 in mining means that the region and the nation are equally specialized in mining, while an LQ of 1.8 means that the region has a higher concentration in mining than the nation.”

 6 Vehicles were not always listed as tools for some occupations for which other O*NET data show that driving is a frequent work activity. For example, no vehicles are listed as tools used by real estate sales agents, although the work context is scored 93/100 for “in an enclosed vehicle or equipment.” The database was helpful in identifying occupations for which the person drove a vehicle or only operated industrial or construction equipment (e.g., with a listing of “Light Trucks and SUVs”).

About the Author:Jennifer M. Miller is an assistant professor (teaching) in the Sol Price School of Public Policy at the University of Southern California.