Socioeconomic Impacts of Automated and Connected Vehicles (2018)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

25 APPENDIX A: WHITE PAPER Synthesis of the Socioeconomic Impacts of Connected and Automated Vehicles and Shared Mobility Johanna P. Zmud, Texas A&M Transportation Institute, USA Nick Reed, Bosch, United Kingdom ing, computing, or other digital technologies, leading to increased privacy risk. Some examples of CAVSM infor- mation that could identify an individual include credit card transactions, biometric data as well as video data or GPS tracks. A single piece of data can be personally identifiable information (PII), such as an address. Like- wise, multiple pieces of data when merged can be PII, even when the individual pieces would not be. Treatment of PII is distinct from other types of data because it needs to be not only protected but also collected, maintained, and disseminated in accordance with the fair information practices (in the United States) or according to regulation (in the European Union). While the United States and the European Union have privacy frameworks in place, there is no specific legislation or regulations that speak to the ownership and security of personal information gener- ated or transferred by CAVSM. Safety and Security Traffic safety benefits are a fundamental motivator for connected and automated vehicle (CAV) development and deployment. More than 90% of traffic crashes are estimated to be caused by human error. CAV is expected to mitigate crash risk stemming from human error with the potential for significant societal benefits. However, evidence of the safety benefits are still being gathered, pri- marily through public road testing of the vehicles taking place both in the United States and in EU Member States. Not only may CAVs mitigate some errors, but also they may introduce new types of driving and vehicle operation errors. As with CAVs, shared mobility operations have the potential to both mitigate and exacerbate human-error Summary Vehicles that are increasingly connected, automated, and shared have the potential to change personal, freight, and public transportation profoundly. While the transition to widespread adoption of connected and automated vehicles and shared mobility (CAVSM) is underway in the United States (U.S.) and in the Euro- pean Union (EU) Member States, uncertainty exists around the pace, scope, and impacts of the potential end states of the transition. The potential benefits to society are immense. On the other hand, the technolo- gies will solve some problems but could also create new ones. This paper discusses the high level implications of CAVSM on four important socioeconomic issues: data privacy and access, safety and security, economics and workforce, and equity. Key points related to these four topics are presented below. Data Privacy and Access CAVSM is characterized by unprecedented volumes and new types of data. These data are used to improve traffic and vehicle safety, environmental outcomes, and acces- sibility; streamline the movement of people and goods; and bring direct commercial benefit through provision of innovative customized mobility services. Positive socioeconomic outcomes are contingent on adherence to voluntary or regulatory guidelines for data privacy protection and to established protocols for data access and use. CAVSM has the potential to weaken traditional means of protecting individuals’ privacy through its broad reliance on various mobile, sensor, global position-

2 6 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S caused traffic crashes. In terms of the former, shared mobility operations could provide alternatives to driving for some at-risk drivers; in terms of the latter, increased congestion at the curbside due to proliferating pick-ups and deliveries increases potential for crashes among vehi- cles as well as other road users. In addition, as software and connectivity play a much bigger and more critical role for the safe operation of CAVs, these vehicles may be at greater risk for cyber-attacks. Security by design is an approach followed to mitigate cyber risk in both the United States and in the European Union. Economic and Workforce Issues Mobility is directly associated with economic prosperity. Thus, the introduction of CAVSM services could influ- ence the availability, cost, and efficiency of freight and passenger transport services. The movement of goods is often cited as a low-margin activity, and so improve- ments in efficiency through CAV is being aggressively pursued in the freight sector. Automation of the long- haul truck driving task has been estimated to reduce total transportation costs by about a third through 2040. Real-world trials of truck platooning have shown specific improvements in fuel efficiency. CAVSM also is expected to improve passenger transportation opportu- nities for many segments of society. Connectivity and shared transportation can be used to enable greater demand responsiveness, while automation may be able to reduce operating costs. However, these outcomes may lead to induced demand and increases in vehicle travel and emissions if demand is not responsibly man- aged. CAVSM operations also reflect the emergence of the so-called ‘platform’ or ‘gig’ economy, which offers flexibility for workers but may lack social protections. However, it is the potential for automated vehicles (AVs) to reduce employment that is perceived as a key concern. The International Transport Forum has predicted that up to 4.4 million trucking jobs could be eliminated in the United States and Europe; similarly a U.S. study suggests that automation is likely to have significant negative impact on truck drivers, bus drivers, and taxi drivers. It should be noted that automation may also create new employment opportunities. Equity Social equity relates to the fair distribution of services across potential recipients. For CAVSM, the vehicle designs and technologies used, the market segments addressed, and the regulations imposed upon such ser- vices are all factors that influence how mobility benefits will be distributed. For example, the best safety systems are currently being fitted primarily on new luxury vehi- cles. It will take the cascading of such safety systems from luxury to mass-market vehicles for the equitable distribution of safety benefits. Four significant areas where CAVSM might have positive equity impacts— access to employment, access to education, access to health services, and access to discretionary travel for social purposes—support greater societal well-being. An important issue of equity is the extent to which trans- port services enable those with additional travel needs, such as the disabled and/or elderly to satisfy their mobil- ity requirements. Questions of who gets served and at what cost are significant policy issues to guide the proliferation of CAVSM. IntroductIon Key Takeaways • In the United States, automated vehicle (AV) and connected vehicle (CV) systems are viewed as independent technologies, whereas in Europe they are seen as complementary. • Connectivity is seen to be a major enabler for driverless vehicles in the medium term. • AVs can be connected, whereas CVs will not necessarily be automated. • Connected and automated driving facilitates the conditions for shared mobility services, which refers to a business model in which physical assets are accessed sequentially or concurrently by multiple users on a pay-per- use basis. • Coupling the development of new CAVSM to emerging communication standards may delay exploitation of the benefits that CAVSM may offer. The purpose of this paper is to provide foundational information on the socioeconomic impacts of AVs, con- nected vehicles (CVs), and shared mobility, covering the transport of people and goods. When referencing all three of these mobility technologies, the acronym CAVSM is used in this paper. Because CAVSM mobil- ity innovations are developing and proliferating at a rapid pace, there is a need for informed, proactive, and consistent evaluation in the planning, deployment, and assessment of them and their potential socioeconomic impacts. This is important not only now as they operate as independent mobility services, but also in the future

A P P E N D I X A : W H I T E P A P E R 27 as the three technologies integrate in emerging appli- cations, such as CAVs and shared automated vehicles (SAVs). This paper assumes that AVs can be connected, whereas CVs will not necessarily be automated. Descriptions of CAVSM AV technologies represent a switch in responsibility for the driving task from human to machine. They encom- pass a diverse range of automated technologies, ranging from relatively simple driver assistance systems to fully automated (or autonomous) vehicles. The Society of Automotive Engineers (SAE) International has catego- rized the levels of automation into six levels (see side- bar). A highly automated vehicle (Levels 4 and 5) does not require a steering wheel, accelerator or brake pedal. AV driving functionality is handled through onboard computers, software, maps, and radar and lidar sen- sors. Highly automated vehicles are not yet operating freely on public roads (other than as pilot programs). Currently, vehicles available to consumers are primarily Level 1 or 2 automation. Since most passenger and commercial vehicle traffic accidents are caused by “human errors,” the safety ben- efits AVs could provide are compelling–although incon- trovertible empirical proof that AVs deliver safety benefits has yet to be produced. Other potential benefits relate to congestion mitigation, air pollution and greenhouse gas (GHG) reduction, and mobility enhancement for under- served populations, such as low-income people, older adults, the disabled, and rural residents. Supported by advancements in artificial intelligence (AI)–particularly in the areas of Big Data analytics, machine learning and knowledge management–rapid progress is being made in terms of AV development and deployment. A CV has internal devices that enable it to commu- nicate wirelessly with other vehicles, as in vehicle-to- vehicle (V2V) communication, or with an intelligent roadside unit, as in vehicle-to-infrastructure (V2I) com- munication. V2V applications enable crash prevention, and V2I applications enable telecommunication, safety, mobility, and environmental benefits. The acronym V2X is sometimes used to designate vehicle-to-everything (including pedestrian and bicyclist) communication. Data communications that enable real-time driver advi- sories and warnings of imminent threats and hazards on the roadway are the foundation of connected vehicles (Hong et al. 2014). At present, the V2I and V2V appli- cations solely provide driver alerts; they do not control vehicle operations. Dedicated short-range communica- tion (DSRC) and 4G-LTE are two widely used candidate schemes for CV applications, and 5G is on the horizon. In Europe, the term “connected and automated driving” (C&AD, or CAD) refers to a set of systems using sen- sors, AI, and other technologies that enable vehicles to travel without direct human operation and to exchange information wirelessly with other vehicles, infrastruc- tures and third-party service providers (European Com- mission 2017b). In the United States, AV and CV systems are often viewed as independent technologies, whereas in Europe they are seen as complementary. Connectivity is seen to be a major enabler for driverless vehicles in the medium term. C&AD facilitates the conditions for shared mobil- ity services. Unlike CV and AV that refer specifically to technology, shared mobility refers to a business model in which physical assets (e.g., bicycles, automobiles, deliv- ery trucks, etc.) are accessed sequentially or concurrently (e.g., pooling) by multiple users on a pay-per-use basis. This model enables users to obtain short-term access to transportation services as needed and with seamless payment transactions mainly through mobile devices or online platforms (Shaheen et al. 2017a). Shared mobil- ity is an alternative to ownership, or it may complement car ownership in households and conventional public transport. According to McKinsey & Company, Europe and the United States represent two of the three core regions comprising a shared mobility market of nearly $54 billion in 2016 (Grosse-Ophoff et al. 2017). The United States is one of the largest markets at $23 bil- lion and is dominated by ridesourcing, while Europe is much smaller at just under $6 billion and leans more Levels of Automation Level 0: No automation. Level 1: Human controls driving, but the automated systems can take over one major driving function, such as steering or speed. Level 2: Human is responsible for safety- critical functions. Automated systems can execute both steering and acceleration/ deceleration functions to assist driver. Most automakers are currently developing vehicles at this level. Level 3: Vehicle can manage all safety- critical functions under certain conditions, but human is expected to take over driving tasks when alerted. Level 4: Vehicle is self-driving in some condi- tions or situations but not all. Level 5: The car can be completely self- driving in all situations. Requires absolutely no human participation in driving task.

2 8 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S toward carsharing. The convergence of C&AD and shared mobility is known as SAV, and there are various small-scale pilots in the United States and in EU Mem- ber States as discussed later in this paper. Many people believe that highly automated vehicles will first be avail- able to consumers as SAVs (The Economist 2018). Cost is a main factor. Lidar sensors are still too expensive to be used in mass produced vehicles. The cost of this tech- nology is considered less of a barrier for fleet vehicles because they will be generating revenue throughout the day to cover the expense, whereas the typical privately owned vehicle is used for a small fraction of a day. The key enabler for CAVSM is communication of location and status data and an ability to analyze and interpret this data intelligently. While emerging forms of connectivity (e.g., DSRC; 5G mobile communications) offer promise for new communication services, many practical benefits of CAVSM can be achieved over exist- ing mobile networks in the majority of the United States and EU Member States. Coupling the development of new CAVSM to the emergence of emerging communi- cation standards may delay exploitation of the benefits that CAVSM may offer. The next section provides a brief summary of the reg- ulatory frameworks for CAVSM in the United States and in the European Union as a general context to inform the topical discussions that follow. Regulatory Frameworks United States The U.S. Department of Transportation (U.S. DOT) pub- lished a federal automated vehicle policy via the National Highway Traffic Safety Administration (NHTSA) in late 2016 that took initial steps toward a unified, national regulatory framework for AVs. Then a year later, NHTSA issued Automated Driving Systems 2.0: A Vision for Safety that replaced the earlier policy framework (NHTSA 2017). It offered voluntary guidelines for the AV industry in designing best practices for testing and deployment of AV vehicles that incorporate SAE Levels 3–5 or highly automated vehicles. The policy framework did not carry a compliance requirement or enforcement mechanism. Instead, it offered suggestions on priority safety design elements and encouraged industry participants to per- form voluntary safety self-assessments that demonstrate their approach to testing and deployment. It also clari- fied NHTSA versus states’ responsibilities in this area. NHTSA regulates motor vehicles and motor vehicle equipment, while states are responsible for regulating the human driver and most other aspects of motor vehicle operation. In 2018, the U.S. DOT plans to release a third iteration of the guidance, AV 3.0. While the 2017 policy framework was focused on passenger vehicles, the 2018 policy guidance is expected to cover all transportation modes, including public transit, rail, commercial trucks, and aviation. Federal regulatory action for CVs has focused on V2V technology, rather than V2I technology. In August 2014, NHTSA issued an advance notice of proposed rule- making to begin implementation of V2V communications technology. Then in January 2017, NHTSA issued a pro- posed rule to establish new Federal Motor Vehicle Safety Standards to mandate V2V communications for new light vehicles and to standardize the message and format of the V2V transmissions. However, as of 2018 such rule- making has not advanced. In November 2017, NHTSA issued a statement that it has not made any final decision on the proposed rulemaking concerning a V2V mandate. In September 2017, the U.S. House of Representa- tives passed the SELF DRIVE Act, and the U.S. Senate followed by passing the AV START Act in October 2017. As of May 2018, Congressional action has not moved forward toward passage. These acts were in response to calls for regulatory changes at the federal level to promote the development of AV technology. Both acts preserve the existing differentiation of responsibilities between NHTSA and the states. The two acts take different approaches to privacy and cybersecurity. The SELF DRIVE Act provides that a manufacturer may not mar- ket a highly automated AV unless that manufacturer has developed a Privacy Plan and a Cybersecurity Plan that identifies, mitigates, and prevents privacy and cyber- security vulnerabilities. The AV START Act establishes a Data Access Advisory Committee to produce a report to Congress with policy recommendations on owner- ship and control of data generated or stored by AVs. The AV START Act does require that manufacturers have a detailed plan for identifying and reducing cybersecurity risks. State legislatures in the United States are becoming increasingly engaged on the topic of AVs. The National Conference of State Legislatures’ (NCSL) Autonomous Vehicles Legislative Database provides current infor- mation on state legislative efforts (see http://www.ncsl. org/research/transportation/autonomous-vehicles-self- driving-vehicles-enacted-legislation.aspx.). According to NCSL: • Forty-one states and Washington, D.C. have consid- ered legislation since 2012, and • Of those, 22 states and D.C. have passed legislation. The states’ legislation has been varied. Some states only enable testing, while other states enable use of an automated driving system on public roads and require a human driver should be in the test vehicles. A few

A P P E N D I X A : W H I T E P A P E R 29 states have recently updated their legislation to remove requirements that a human driver should be behind the wheel at all times. The legislation has been state-specific with no attempt at coordination across states, prompt- ing the congressional action discussed previously that attempts to provide a national policy framework. As the technology for AVs continues to develop, state leg- islation will continue to evolve to address the potential impacts of these vehicles on the road. In terms of a regulatory framework for shared mobil- ity, state legislatures have been involved in regulating Transportation Network Companies (TNCs) and car- sharing programs. As for other types of shared mobility, e.g., bikesharing, these are governed by local government regulations. For TNCs, as of August 2017, 48 states and Washington, D.C., have passed at least one piece of leg- islation regulating some aspect of TNCs (Moran et al. 2017). The amount and degree of regulation varies from state to state: • Forty-three states and D.C. have laws that address operating permits and fees, background check require- ments, operational standards, and protections for passengers. • Five states have laws that address only insurance requirements for TNCs and TNC drivers. A majority of state legislation includes preemption of the local authority to regulate, tax, or impose rules on TNCs. According to NCSL, a handful of states have enacted carsharing legislation. The legislation covers such issues as incentives to use carsharing, carsharing taxation, electrification of carsharing fleets, and creating a regulatory framework for peer-to-peer carsharing (see http://www.ncsl.org/research/transportation/car-sharing- state-laws-and-legislation.aspx). European Union In May 2018, the European Commission issued a Com- munication on an “EU strategy for mobility of the future” to harmonize the legal framework, research, and industrial innovation across Member States (European Commission 2018). In this strategy document, the Euro- pean Commission put forth a progressive and harmo- nized approach to regulation of connected and automated mobility based on experience gained through demonstra- tions and large-scale testing to validate the safety of the technologies. It identified relevant automation use cases: • Passenger cars and trucks at Levels 3 and 4 that are able to handle specific situations on the motorway (e.g., truck platooning convoys) and some low-speed situa- tions in cities (e.g., valet parking) available by 2020. • Public transport vehicles at Level 4 able to cope with a limited number of low-speed driving situations (e.g., urban shuttles for dedicated trips or small delivery vehicles) by 2020. • The European Commission is linking policy and regu- latory initiatives around these use cases. In addition the European Commission provides funding to sup- port demonstrations and large-scale testing through Horizon 2020. In addition, it will provide support in 2018 for testing the use of 5G connectivity to enable highly automated driving functions and new mobility services. In the just-issued communication, the Euro- pean Commission will intensify coordination with/ among Member States so that traffic rules can be adapted to automated mobility in a harmonized way, such as with the 1949 Geneva Convention and the 1968 Vienna Convention on Road Traffic. As part of a revision of the General Safety Regulation for Motor Vehicles, the European Commission is also proposing to regulate: • Data recorders for AVs to clarify whether the vehicle or a driver was in control during an accident, • Platooning to ensure standardization of data exchange across different technologies, and • Protection of vehicles against cyber-attacks. The current European Commission strategy builds upon recommendations of the high level group (GEAR2030) that emphasized the need for a harmonized and cross- border regulatory framework for testing, communication, data security, safety, and cybersecurity (Government of Netherlands 2016). This document indicated that Mem- ber States will rely on a voluntary commitment of the industry to include connectivity in all new vehicles from 2019 onward. Therefore, no mandatory V2V or V2I regulation was envisioned. Member States are also individually moving forward with regulation. There is a challenge to implement an EU-wide legal system considering the divergence of approaches among some Member States. For example: • In 2016, France launched a decree regarding the test- ing of C&AV on public roads, which specified that by 2020, official standards to regulate tests would be operative. • In 2016, Finland created a system of test plates and protocols for automated vehicle trials issued by the national transport safety agency, Trafi.

3 0 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S • Under a broader digitalization initiative, in 2017, Estonia made it legal to test self-driven vehicles on all national and local roads in the country. • Germany has no specific legal framework for the testing of automated vehicles, but testing in traffic is allowed with special permission. • In Spain, national authorities have published a legal framework for public road testing that entails spe- cific requirements for the application and granting of authorization for automated vehicle tests and trials on public roads. • The United Kingdom (UK) is currently conducting a law review that includes the allocation of civil and criminal responsibility by law where there is shared control between humans and computers; the role of automated vehicles in public transport, carsharing and on-demand passenger services; the impact on other road users and how they can be protected from risk; and determining who the responsible person is in a self-driving vehicle. However, the Member States have agreed to work transparently on the development of national legislation affecting consistent EU-wide deployment of C&AD. With a goal of consistency, the focus is on the role of the driver, the transfer of control from human to machine, and traffic behavior. Related, many Member States (except Spain and the UK) are signatories of the Vienna Convention, which makes it mandatory for a driver to be able to control the vehicle (Article 8). New amendments came into force in March 2016 (ETSC 2016a). The key amendment allows a car to drive itself, as long as the sys- tem “can be overridden or switched off by the driver.” A driver must be present and able to take the wheel at any time. The interpretation in Member States’ traffic codes has to still be adapted to enable Level 3—conditional automated driving. In Sweden, new legislation for trials has been proposed that enables testing on public roads as long as the manufacturer takes the responsibility. While the European Commission has great interest in promoting sustainable urban mobility, such as different variations of shared mobility services, there is a frag- mentation of responsibilities among local, regional, and national entities (Gudmundsson 2013). The European Commission has indirect tools at its disposal, either via the Member States or via the so-called soft-law. For example, in its announced Urban Mobility Package, the European Commission has requested the establishment of voluntary Sustainable Urban Mobility Plans, which will serve as a comprehensive planning tool for cities in the areas of land use, road charging and emission reductions, among others. It lacks enforcement power against those who will not comply with the plan. Technically, the European Commission could choose to impose mandatory measures on the Member States, which in turn will have to mandate and regulate cities. But this would raise significant governance challenges. Over the past several years, the development of new technology has drastically changed how society func- tions. Mobile smartphones and online social networks are prime examples of technologies that have become ubiquitous in many people’s lives. While these technolo- gies have become invaluable to their consumers and citi- zens, they have also created a host of new data privacy and access challenges. A similar dynamic is playing out in the transportation sector in terms of CAVSM tech- nologies. The next section highlights some of the impor- tant issues. TABLE 1 Key Aspects of U.S. and EU Regulatory Frameworks for CAVSM United States European Union 2017: NHTSA issued voluntary guidelines for the AV industry in best practices for testing and deployment of highly automated passenger vehicles. 2018: European Commission communication to harmonize the legal framework, research, and industrial innovation across Member States. NHTSA regulates motor vehicle equipment, while states regulate the human driver and motor vehicle operation. European Commission is linking policy and regulations to use cases: (1) Levels 3–4 passenger cars and trucks on motorways and in cities, (2) Level 4 public transport vehicles in low-speed situations—both by 2020. NHTSA rulemaking on V2V mandate has not advanced; neither has Congressional action in the form of SELF DRIVE Act (House) and AV START Act (Senate)—all in 2017. 2018: European Commission is providing support for testing 5G to enable highly automated driving and new mobility services. Voluntary commitment of industry to include connectivity in all new vehicles. Since 2012, 22 (of 50) states and D.C. have passed automated vehicle legislation pertaining to testing and use on public roads. Legislation is state-specific, not harmonized. European Commission intensifying coordination among Member States to harmonize traffic rules for automated mobility. EC proposing to regulate data recorders for AVs, platooning, and protection against cyber-attacks. Forty-eight states have passed at least one piece of legislation regulating some aspect of shared mobility services. Fragmentation of responsibilities for shared mobility among local, regional, and national entities.

A P P E N D I X A : W H I T E P A P E R 31 data PrIvacy and acceSS for color identification; lane departure, read collision, and pedestrian alerts; and a lidar sensor on the roof used for generating a 3D map of the environment (Bloom et al. 2017). These sensors capture continuous data about the vehicle itself as well as the surround- ing environment (i.e., people, vehicles, infrastructure within it). • Most shared mobility services rely on smartphone apps or online platforms to connect paying travelers with the mobility fleets. Payment is often managed through the app or online platform, which stores credit card information. These services also have access to mas- sive amounts of data on both the transport network (such as the current levels of speed and congestion), and on their passengers or clients (such as access/egress locations, routes taken, time of day and frequency of travel). The data are used for the internal optimization of the shared services, but they are also increasingly shared with third parties (Franckx 2017). Moreover, shared services can set up partnerships with cities and transport authorities in which data are integrated and shared for specific public services. The unprecedented volumes and new types of data gen- erated by CAVSM have the potential to improve safety, environmental outcomes, and accessibility; streamline movement of goods and people; and bring direct commer- cial benefit based on enhancing the consumer experience. Realizing these societal benefits, however, is contingent upon addressing data privacy and access issues. Data Privacy Data privacy is defined as the capability of individuals to “determine for themselves when, how, and to what extent information about them is communicated to others” (Westin 1967). This is particularly relevant to privacy of PII, which are any data that could potentially iden- tify a specific individual, including any information that could be used to distinguish one person from another or that could be used for de-anonymizing anonymous data. There is no one list of what constitutes PII. Some examples of information that could identify an individual include name, address, date and place of birth, and bio- metric data as well as video data or GPS tracks of daily mobility. A single piece of data can be PII, such as a home address. Likewise, multiple pieces of data when merged can be PII, even when the individual pieces would not be. Radical transformation of computing, mobile, sensor, global positioning, and database technologies have weak- ened traditional means of protecting individuals’ pri- vacy, leading to increasing risks associated with misuse of PII. Treatment of PII is distinct from other types of data because it needs to be not only protected but also Key Takeaways • CAVSM data enable individuals to be located in specific space and time. The more detailed the spatial location, tem- poral position, or individual information included in the data, the more privacy sensitive the data are and the greater the privacy risk. • Innovations in computing, mobile devices, sensors, and global positioning systems have weakened traditional means of protecting PII. • There are varying models of data access ranging from greatest ease of use (i.e., open access) to greatest privacy protection (i.e., restricted access). • In the United States, there is no single comprehensive legislative framework for data privacy protection; instead, privacy protection relies on fair information prac- tices. In the EU, data privacy is akin to a constitutional right. Why Data Privacy and Access Are Important Socioeconomic Impact Issues for CAVSM The paper’s introduction defines and distinguishes the mobility technologies that are the focus of the EU-U.S. Symposium. The characteristics that they have in com- mon are the collection, transmission, and application of large volumes of data. • CVs receive and share data from onboard computers and sensors with manufacturers, other vehicles, other road users, infrastructure, and third-party service pro- viders. These data relate not only to vehicle opera- tions on the road and in-vehicle diagnostics, but also to users and their personal requirements. Data include location, driver behavior, biometrics, vehicle health, fuel consumption, vehicle emissions, personal com- munications, and infotainment selections (US Govern- ment Accountability Office 2017). • AVs require extensive data to operate effectively. Their sensors and systems typically include: GPS for naviga- tion; a wheel encoder for monitoring the movements of the car; radar on the front and rear bumpers for identifying traffic; a camera near the rear-view mirror

3 2 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S collected, maintained, and disseminated in accordance with the fair information practices (in the United States) or according to regulations (in Europe). Balancing agen- cies’ needs for using such data with individuals’ concerns about their data privacy is a complicated challenge. For example, according to a 2015 survey by the Pew Research Center, a majority of Americans believe it is important—often “very important”—that they be able to maintain privacy and confidentiality in commonplace activities of their lives (Madden and Rainie 2015). Most strikingly, these views are especially pronounced when it comes to knowing what information about them is being collected and who is collecting it. These feelings also extend to a desire to maintain privacy when moving around in public. Survey results from early 2015 show that 63% felt it was important to be able to “go around in public without always being identified.” All adults, regardless of age or gender, express comparable views. Likewise in a 2015 survey to capture attitudes of EU citizens about issues surrounding data protection, two- thirds of respondents were concerned about not hav- ing complete control over the information they provide online (European Commission 2015). A majority were concerned about the recording of their activities via payment cards and mobile phones (55% in both cases). Most do not trust landline or mobile phone companies and internet service providers (62%) or online businesses (63%). Even though people are worried on their privacy, there is often an inconsistency between people’s attitudes about privacy and their behaviors vis-à-vis social media and other digital platforms. Data Access Data access is directly associated with data privacy. Data access refers to a user’s ability to retrieve data stored within a database or other repository. Entities that have data access can move, use, or manipulate the stored data. Rules for accessing data are critical in a Big Data environment because traditional approaches for privacy protection via informed consent and de-identification may no longer be effective (Kum and Ahalt 2013). • Informed consent refers to permission granted by a person to participate in a data gathering activity with full knowledge of the possible risks and benefits of that participation. True informed consent is impos- sible when data are not knowingly provided by a per- son but result from an opportunistic sensing system. For example, how does one provide notice to individuals whose data have been collected via roadside Bluetooth® sensors? • De-identification is a general term for any process of removing the association between a set of identifying data and the data subject (Garfinkel 2015). The term is often used interchangeably with anonymization. It attempts to balance the contradictory goals of using and sharing information about people with protecting their privacy. In recent years with increasing imple- mentation of data science analytics, researchers have shown that de-identified data can often be re-identified through linkages among multiple datasets. Kum and Ahalt (2013) identify varying models of data access that range from greatest ease of use to great- est privacy protection. These models are: open, moni- tored, controlled, and restricted. Examples of each of the models in CAVSM applications are noted below. • Open access: Data are freely available online to all at no cost with limited restrictions as to reuse. Data are typically sanitized (i.e., standard disclosure limi- tation methods are applied) to allow public access. An example is advanced apps that employ open data, algorithms, and advanced programing interfaces (known as APIs) to aggregate real-time information services, multi-modal trip planning and fare payment into a single application, such as the Open Mobility Project in Berlin (see https://blog.bosch-si.com/mobility/ intermodal-transportation-to-advance-mobility-in- urban-areas/). • Monitored access: Access to data typically requires some type of user authentication, and the data are usu- ally aggregated. An example is Uber Movement data where Uber makes its trip data available via a public website to users who request and receive approval to access it (see https://movement.uber.com/?lang=en-US). • Controlled access: Access to data is controlled through the use of specialized software or a specialized plat- form. For example, specialized software is necessary to retrieve and analyze the data stored in Event Data Recorders (EDRs) (i.e., a vehicle’s “black box”) (Koch 2006). • Restricted access: Access to information is restricted through decoupling, meaning that PII is separated out from sensitive data. An example of how vehicle data can be restricted is through blockchain technology. A simple description of blockchain may first be neces- sary. A “block” is a record of new transactions (e.g., a vehicle location, a mile traveled, a trip taken). Once a block is completed it is added to the “chain,” creating a chain of blocks. The blockchains are interconnected such that each subsequent block contains a crypto- graphic image of the previous block. Thus data cannot be changed without recognition of that fact after the respective data have been entered into a block, com- pleted, and “attached” to a subsequent block (Dorri

A P P E N D I X A : W H I T E P A P E R 33 et al. 2017). Every block in the chain is linked to a published public key that represents a particular user. That key is encrypted so that the user cannot be identi- fied. CAVSM use cases could include carsharing, ride- sourcing, or CV-enabled road pricing schemes. As described in the following section, the United States and Europe follow differing regulatory frameworks in terms of data privacy protection and data access control. Privacy Protection United States The United States has a patchwork of federal and state laws and regulations that overlap, dovetail, and may even contradict one another (Jolly 2017). At the federal level, different privacy requirements apply to different industry sectors (e.g., health or financial information). The laws are often narrowly tailored and address specific data uses and users. An example is regulation pertaining to EDRs. EDRs store information produced immediately before and during an accident, such as date, time, vehicle and engine speed, steering angle, throttle position, brak- ing status, force of impact, seatbelt status, and air bag deployment. None of these data elements are PII, but when combined with other technologies, such as onboard navigation systems or mapping apps, EDR data could be used to personally identify an individual (Canis and Peterman 2014). The Driver Privacy Act of 2015 pro- vides that all car manufacturers must install EDRs, and all EDRs must collect specific information. It also stipu- lates that the EDR information belongs to the owner or, in the case of a leased vehicle, the lessee of the vehicle in which the EDR is installed. EDR data are restricted, accessed (via specialized software) and are shared only with the consent of the vehicle owner or lessee. For those entities not subject to industry-specific regulation, the Federal Trade Commission (FTC) is the primary federal privacy regulator (Sotto and Simpson 2014). It uses Section 5 of the FTC Act, which is a gen- eral consumer protection law that prohibits “unfair or deceptive acts or practices in or affecting commerce” to bring privacy enforcement actions. Yet, in general, FTC enforcement has been mostly procedural, focusing on companies’ notice and consent actions, such as ensur- ing that online companies have privacy policies, that the policies are not hidden in obscure places on company websites, etc. Most states have enacted some form of privacy legis- lation. However, California leads the way in the privacy arena, having enacted multiple privacy laws, some of which have far-reaching effects at a national level, such as California’s Confidentiality of Medical Information Act. Unlike many federal privacy laws in the United States, California’s privacy laws resemble a European proactive regulatory approach to privacy protection. However, even in California, there is no regulatory framework that specifically addresses CAVSM data. Instead, there are many guidelines developed by governmental agen- cies and industry groups that are not legally enforceable but are part of self-regulatory efforts that are considered best practices in the context of CAVSM. The automotive industry developed privacy principles in 2014 largely in response to data privacy and security concerns raised by U.S. Congressional members about the increasing connectivity and automation of auto- mobile technology (Markey 2015). The auto industry privacy principles, effective for new vehicles manufac- tured no later than model year 2017, represent a uni- fied response to such concerns (Alliance of Automobile Manufacturers and Association of Global Automakers 2014). Overall, the privacy principles require clear and prominent notices about the collection of information, the purposes for which it is collected, and the types of entities with which the information is shared. Europe Unlike in the United States, the right to privacy is a highly developed area of law in the EU. Until May 2018, the pro- cessing of personal data was regulated by the Data Protec- tion Directive. This was an EU Directive adopted in 1995 that identified conditions under which personal data may be processed—transparency, legitimate purpose, and pro- portionality. This Directive has since been replaced by the General Data Protection Regulation (GDPR) that was approved by the EU Parliament in 2016, and it is subject to enforcement as of May 25, 2018. It is important to note that GDPR is a Regulation and not a Directive. A regulation is a binding legislative act. It must be applied in its entirety across the EU, while a directive is a legislative act that sets out a goal that all EU countries must achieve. However, it is up to the individ- ual countries to decide how to achieve the goal. Thus, the GDPR serves to harmonize data protection regulation across Member States (see https://www.eugdpr.org/key- changes.html). Its main goal is protection against privacy and data breaches. It covers “personal data” which is any information that can be used to directly or indirectly identify the person. Key provisions include the following: • It applies to all companies processing personal data of data subjects residing in the EU Member States regardless of the company location or where the pro- cessing takes place. • The request for consent must be given in an intelligible and easily accessible form, with the purpose for data processing attached to that consent request. • Breach notification is mandatory, within 72 hours of awareness.

3 4 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S • Data subjects have the right to obtain confirmation as to whether personal data concerning them is being processed, where and for what purpose, as well as a copy of their personal data that is provided free of charge and in electronic format. They can also request that their data be erased and that processing of it cease. • Privacy by design, which is a framework based on pro- actively embedding privacy into the design and opera- tion of information technology (IT) systems, networked infrastructure, and business practices from the start of systems design, is a legal requirement. • Breach of GDPR requirements can be fined up to 4% of annual global turnover or 720 million (which- ever is greater). While the United States and the European Union have privacy frameworks in place, there is no specific legisla- tion or regulations that speak to the ownership and secu- rity of personal information generated or transferred by CAVSM. As vehicles become increasingly connected, auto- mated, and shared, so the volume of data they collect, combine, store and communicate increases. Complex questions arise as to whether such data constitutes “per- sonal data” and, if so, who is responsible for it and how is it secured. While not all data collected by CAVSM will on its own identify an individual driver, passenger or user, in many cases it may be combined with other infor- mation to identify such individuals, and therefore it may be “personal data.” For example, in the EU, location data collected by smartphones is generally considered to be personal data because individuals can be directly or indirectly identified through their patterns of movement. By analogy, geo-location data collected by CAVSM is likely to be considered personal data where this data alone or in conjunction with other information identi- fies an individual driver, passenger or user through their patterns of movement. The importance of the type of regulatory approach a nation follows is significant when considering that data emanating from connected, auto- mated or shared vehicles constitutes PII. CAVSM and Privacy Risk CAVSM data enable individuals to be located in a specific space and time. The more detailed the spatial location, temporal position, or individual information included in the data, the more privacy sensitive the data are and the greater the privacy risk. Privacy risk is defined as a func- tion of “the likelihood that a data action causes prob- lems for individuals, such as loss of trust or economic loss, and the impact of the problematic data action” (Brooks and Nadeau 2015). Collection, retention, logging, generation, transformation, disclosure, and transfer are examples of data actions. One potentially problematic data action, for example, is surveillance in which personal data are used to track the activities and whereabouts of an individual in a way that may not be proportional to the service being provided. Some have suggested that an AV’s sensors that scan the surrounding environment while operating on public roads equates to surveillance activity (Bloom et al. 2017). Two criteria are usually applied as a means of analyzing and categorizing use cases according to their privacy risk: likelihood of a problem and magnitude of harm (Zmud et al. 2016a). Criteria 1—Likelihood of a Privacy Problem The likelihood of a privacy problem occurring is the probability that a data action will generate a problem for the typical individual whose personal information is processed. Various factors associated with a particular use, as noted in Figure 1, will impact the probability of a privacy problem occurring. Uses of CAVSM data that enable real-time applica- tions raise fewer privacy concerns because personal data are not central to the use. In contrast, when data are retained or stored (instead of deleted) to analyze behav- ior, the privacy risk increases because the sensitive data could be involved in a problematic data action. Stored data simply allows more time for the data to be dis- closed through an intentional or accidental data action. Second, if recurrent information about an individual’s actions over time are amassed, that information may be used to track a person’s whereabouts and activities. Both of these situations increase the probability of pri- vacy issues. Other factors include the government versus third-party ownership of data, and the geographic com- prehensiveness of the database. Criteria 2—Magnitude of Harm from Privacy Problems Privacy risk is a function of the magnitude of harm a data action creates, multiplied by the likelihood that the problematic data action occurs. The harm, or loss TABLE 2 Key Aspects of U.S. and EU Privacy Regulation United States European Union At federal level, data privacy protec- tion provided by FTC consumer protection law that prohibits unfair or deceptive practices. Data privacy akin to a constitutional right. Most states have some form of privacy legislation; none address CAVSM data. General Data Protection Regulation (GDPR) harmonizes data protection regulation across Member States.

A P P E N D I X A : W H I T E P A P E R 35 incurred, due to a privacy problem may not always be straightforward to quantify. A data action that leads to financial losses such as credit card fraud, can be quanti- fied in monetary terms. However, other losses may be ambiguous as agencies try to consider issues such as the effect of leaking embarrassing activity of individu- als, variation of individual perceptions of privacy risk, and loss of public trust. The magnitude of harm from a potential privacy risk increases as CAVSM data are linked to other data sources (Figure 2). There are three issues associated with CAVSM data that are related to greater or lesser privacy risk. These issues are open data, data sharing, and data ownership. Open Data Open data increases the likelihood of a privacy problem as well as the potential magnitude of harm. Open data is a concept that implies that data should be available to be freely used, re-used, and redistributed by anyone (see http://opendatahandbook.org/guide/en/what-is-open- data/). Many shared mobility platforms, such mobility- on-demand (MOD) rely on the availability open data. The value of open data is that it can be freely intermixed with other “open” material for an enhanced ability to combine different datasets together in order to develop more and better products and services. While some stakeholders call for open data in the interest of research and development across industries or public acceptance of connected and automated technologies, others are pursuing strategic partnerships or turning proprietary data into a business opportunity. The availability and flow of data becomes particularly important where that flow might enhance public safety or other interests. In the United States, the Obama White House signed an Executive Order in 2013 making open and machine- readable the new default for government information with the goal of increasing citizen participation in gov- ernment, creating opportunities for economic develop- ment, and informing decision making in both the private and public sectors. In 2014, the European Commission issued a directive establishing an open data policy in Europe (see https://joinup.ec.europa.eu/sites/default/ files/document/2014-02/EU%20Open%20data%20 policy.pdf). This policy stipulates that all publicly funded data must be available for all and must be easily com- bined with other types of open data (e.g., geo, traffic, tourism) to benefit EU-wide services and applications. In both the United States and Europe, the focus is on non-privileged data, that is, data which do not contain law enforcement information, national security infor- mation, personal information, or the disclosure of infor- mation that is prohibited by law. However, it should be noted that open data are mostly Big Data, whose value is increased through reuse, re-purposing, and linking to other sources. So open data increases the likelihood of a problem occurring and the risk for greater magnitude of harm. Open data standards are critical to ensuring privacy protection. Data Sharing Data sharing increases the likelihood of a privacy problem as well as the potential magnitude of harm. Because the value of data is maximized when different data sources are integrated, data sharing is becoming critical practice for both public- and private-sector agencies. For instance, private ridesourcing companies collect granular data (e.g., exact volume, time of day, O/D, length, speed) that can inform important urban and regional transportation planning or modeling issues. But, from the private-sector perspective, sharing this level of detail might jeopardize not only the individual’s privacy but also the firm’s busi- ness practices and intellectual property. Balancing these factors is a distinct challenge to public-/private-sector data sharing. FIGURE 1 Relative likelihood of privacy problems. Likelihood of ProblemLow No personal information; data not stored Data may be retained or stored Data may be stored and amassed over many observations High FIGURE 2 Relative magnitude of harm from privacy problems. Magnitude of HarmLow Data is not linked to other sources Data may be linked to other data sources Data may be linked to sensitive personal information High

3 6 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S Data-sharing issues will only grow in importance as new transportation service models grow in stature. Again, drawing on the ridesourcing example, how does a local government in the United States adequately regu- late this new industry without working with the same data as the regulated firm? As new models of service provision, such as CAVSM, appear on the horizon, the ability of public agencies to perform their regula- tory roles is called into question unless all sides agree on common data-sharing principles. A possible solution is data sharing via a data exchange, such as the World Bank’s OpenTraffic project (Zipper 2018). It was ini- tially developed as a way to aggregate traffic informa- tion derived from commercial fleets. In 2017, the project became a part of SharedStreets, a collaboration between the National Association of City Transportation Offi- cials (NACTO), the World Resources Institute, and the OECD’s International Transport Forum to pilot new ways of collecting and sharing a variety of public and private transport data. It goes beyond open data and is attempting to develop ways for working with privacy sensitive data, such as collecting aggregated data that is rich enough to allow for deep analysis while still hiding information about individual rides. Still there are challenges in incentivizing private-sector partners like Google, Uber, Lyft, Didi Chuxing, Ofo, and Mobike to participate. Without such a majority of entities in a city or region participating, data availability will affect not only the quantity but also the quality of information that is available for more and better mobility products and services. Data Ownership Data ownership determines whether a privacy problem is likely to happen. Ownership of data is tantamount to control, determining who can collect, process, use, and disseminate data. Ownership also implies who can profit from what is owned. CAVSM data hold significant monetization potential, whether it is vehicle diagnostics data (like speed, tire pressure, etc.) or data regarding customer opinions and driving experiences. For exam- ple, McKinsey & Company has estimated that the car data market could generate as much as $750 billion in revenues by 2030 (Alonso Raposo et al. 2018). Ownership is straightforward when applied to a house since there is a formal transaction with written acknowledgment that makes ownership clear. However, when applied to data, ownership becomes complicated. There are many roles with which the notion of owner could be associated, from the data creator, to the data packager, to the data subject. However, just as impor- tant, ownership implies a broader responsibility—data stewardship—where the owner must consider the conse- quences of how the data are used, particularly for how a particular use might impact data privacy. In the United States, the concept of data stewardship is rooted in a rather loose approach to data governance that solely reflects fair information practice as defined by the FTC (Diamond et al. 2009). In Europe, on the other hand, the GDPR specifically sets requirements on organiza- tions’ data governance and enforces these requirements with financial sanctions. These requirements include issues of data quality and assurance of that quality. The key requirements of GDPR’s Article 5 involve appro- priate usage, accuracy and data security. Specifically, it mandates that “every reasonable step must be taken to ensure that personal data that are inaccurate, having regard to the purposes for which they are processed, are erased or rectified without delay.” Under GDPR, it will be important to do validation at both the time of data collection/entry and at time of use. Safety and SecurIty Key Takeaways • The “traffic crash” externality reflects the social cost of driving that are costs inflicted on fellow road users and spillover effects on the rest of society (such as congestion costs). • Drivers, vehicles, and environmental con- ditions can all cause crashes. However, human errors are a critical cause of more than 90% of crashes. • In the United States CV applications are mostly seen as bringing safety benefits. In Europe, environmental and traffic flow benefits are also cited, and V2I technolo- gies are viewed as fundamental to smart mobility applications. • The more miles/kilometers that AVs travel on different roads, in different environ- ments, and under various weather con- ditions, the more quickly their safety improves and their capability to monitor the surrounding environment increases. • Trust in AV technology is a barrier to acceptance and use. • While cybersecurity issues are a challenge for CVs, security becomes a bigger concern with Level 4 and Level 5 AVs, in which soft- ware and connectivity play a much bigger and more critical role for safe driving.

A P P E N D I X A : W H I T E P A P E R 37 This chapter discusses two inter-related cross-cutting issues: safety and security. Safety often refers to road traffic safety, which is defined as the reduction in harm (deaths, injuries, and property damage) resulting from collisions involving vehicles and/or people traveling on public roads. The most common measures to define road safety are the number of road crashes, the number of road casualties, and the associated negative consequences (Wegman 2017). Traffic safety benefits are a fundamen- tal motivator for CAV development and deployment. Closely related to safety is the topic of security. Up until recently, vehicle security was related to anti-theft or hijacking measures. But current interest in security stems from the convergence between automotive technology and computer technology that has increasingly changed the methods by which motor vehicles are developed and are driven. The introduction of telematics, connectiv- ity, and the integration of smartphones and Bluetooth devices makes vehicles vulnerable to cyber-attacks (IEEE 2018). There is also concern about the security of per- sonal data collected and stored in shared mobility data- bases, as discussed in the previous chapter. Why Safety Is an Important Socioeconomic Impact Issue for CAVSM When people drive a vehicle, they not only increase their own risk of a crash, but also increase crash risks for other motorists, as well as pedestrians and bicyclists. This consequence of driving is known as the “traffic crash” externality. It reflects the social cost of driving, which is conceptually different from the private costs individuals may incur, such as injury, death, or damage costs. Motorists can internalize these private costs by refraining from driving, exercising greater care while driving, or insuring themselves (and vehicles) against possible damages (Jansson 1994). But some traffic crash costs are not internalized by the motorist. There are costs inflicted on fellow road users and spillover effects on the rest of society (e.g., congestion costs, net output losses, and hospital treatment). In such cases, the total costs of the crash are not borne just by the individuals involved (Edlin and Karaca-Mandic 2006; Parry et al. 2007; Anderson et al. 2014). In 2016 in the United States, there were 37,461 people killed in motor vehicle crashes, an increase in lives lost from 2015 and 2014 (respectively, 35,092 and 32,657) (NHTSA 2016). Crash risks are not limited to occu- pants or operators of motorized vehicles. Of the more than 2 million roadway injuries in the United States in 2011, 69,000 were pedestrians and 48,000 were bicy- clists (Anderson et al. 2014). NHTSA estimated the total social cost of motor vehicle crashes in the United States in 2010 as US$242 billion (NHTSA 2015a). The cost components included productivity losses, property damage, medical costs, rehabilitation costs, congestion costs, legal and court costs, emergency services, insur- ance administration costs, and employer costs. In Europe, unlike in the United States, road fatalities are declining. In 2016, 25,500 people were killed. The European Commission estimated the social cost of these road fatalities and injuries to be at least €100 billion (Traffic Impact Newswire 2016). The 2016 fatality esti- mate was 600 fewer than in 2015 and 6,000 fewer than in 2010, and it represented a 19% reduction over the last six years (European Commission 2017a). However, not all Member States have had improvements in road safety since 2010. The countries with the lowest fatality rate per million inhabitants were Sweden, the UK, the Netherlands, Spain, Denmark, Germany, and Ireland. Causes of Traffic Crashes Drivers, vehicles, and environmental conditions can all cause crashes. However, human errors are a critical cause of more than 90% of crashes at the national level in the United States (NHTSA 2015b).1 While a compara- ble statistic could not be found for the European Union in aggregate, the U.S. statistics were often cited and applied to a European context (see for instance, https:// ec.europa.eu/transport/themes/its/road_it). The attribu- tion of critical reasons by NHTSA are presented below and are assumed to apply reasonably well to Europe: • Drivers, 2,046,000 crashes (94%); • Vehicles, 44,000 crashes (2%); • Environment, 52,000 crashes (2%); • Unknown, 47,000 crashes (2%). The driver-related “errors” are broadly classified into: recognition (41%), decision (33%), performance (11%), and non-performance (7%) errors. • Recognition errors include those related to a driver’s inattention, internal and external distractions, and inadequate surveillance. Such errors would include the broad category of distracted driving (NHTSA 2015a). • Decision errors include driving too fast for conditions or too fast for curves, and making false assumptions of others’ actions or illegal maneuvers. Alcohol involved crashes involve both impaired judgment (decision errors) and perception problems (recognition errors). 1 The “critical cause” is defined as the immediate reason for the pre-crash event as collected in NHTSA’s National Motor Vehicle Crash Causation Survey, conducted from 2005 to 2007.

3 8 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S • Performance errors include overcompensation, poor directional control, etc. • Sleep (or drowsy driving) was the most common critical reason among non-performance errors. In NHTSA’s National Motor Vehicle Crash Causation Survey, vehicle-related factors were identified primarily as problems with: tires, brakes, steering column, etc. Environment-related causes were defined as roadway or atmospheric conditions. Impacts of CAVSM on Traffic Crashes CV Technologies Safety messages provided by V2V and V2I technolo- gies should enable drivers or automated vehicle systems to take actions that could reduce the severity of traf- fic crashes or avoid them. Such messages simply warn the driver (in the case of non-highly automated CVs) when there is high risk for collision but do not automati- cally apply the brakes. Their effectiveness depends upon drivers having the applications in their vehicles, turning them on, and paying attention to the warnings. Much of the early evidence about effectiveness of V2V or V2I applications in mitigating traffic crashes is from computer-based simulations. Najm and others (2010) found V2I systems bring only small marginal benefits to the safety benefits of V2V systems alone. The U.S. Gov- ernment Accountability Office (2013) pointed out that organizations researching the benefits of V2V or V2I have noted that the benefits depend on a variety of factors, including the size and location of the deployment, the number of roadside units deployed, the types of appli- cations that are deployed, and that some applications require a majority of vehicles on the road to be equipped before reaching optimum safety benefits. In the United States, development and testing has shifted in the last several years to focus more on V2V applications. This was to facilitate the implementation of safety technologies that do not require state and local governments to make costly infrastructure investments. Also, this shift was in response to what was considered to be impending NHTSA rulemaking on V2V. As noted earlier, NHTSA has since delayed its decision. But at the time, NHTSA and FHWA through the ITS/JPO Joint Pro- gram Office began focusing on evaluating the technical strengths and weaknesses of V2V. A study of V2V devices installed as part of the Connected Vehicle Safety Pilot Model Deployment in Michigan found that the devices were technically able to transmit and receive messages, and safety applications enabled by these devices were effective in mitigating potential crashes (Harding et al. 2014). But it also noted that various aspects still needed further investigation including: the impact of spectrum sharing, ability to mitigate V2V communication conges- tion, incorporation of GPS positioning to improve relative positioning, remedies to address false positive warnings, and driver-vehicle interface performance. More extensive evaluative data on the effectiveness and benefits of specific applications is expected from the Connected Vehicle Pilot Deployment programs in New York City, Tampa, and Wyoming that are cur- rently underway and expected to be completed by 2021. These pilots will also assess the potential negative con- sequences of safety warnings, such as driver distraction. • The New York City pilot aims to improve the safety of travelers and pedestrians by testing and evaluating V2V and V2I vehicle applications and V2I pedestrian applications. The pilot will equip taxis, Metropolitan Transportation Authority buses, United Parcel Ser- vice vehicles, NYCDOT fleet vehicles, NYC Depart- ment of Sanitation vehicles, and pedestrians (see https://www.its.dot.gov/factsheets/pdf/NYCCVPliot_ Factsheet_020817.pdf). • The Tampa pilot aims to improve the safety and mobility of automobile drivers, public transit riders, and pedestrians by also testing and evaluating V2V and V2I vehicle applications and V2I pedestrian appli- cations. This pilot will equip privately owned vehicles, buses, streetcars, and pedestrians (see https://www.its. dot.gov/pilots/tampa_participants.htm). • The Wyoming pilot aims to improve driver safety along Interstate 80 by testing and evaluating V2V and V2I applications that provide advisories, road- side alerts, and dynamic travel guidance. The pilot will equip 400 fleet vehicles and commercial trucks (see https://www.its.dot.gov/pilots/pdf/04_CVPilots_ Wyoming.pdf). Unlike in the United States where CV applications are mostly seen as bringing safety benefits, in Europe CV appli- cations are also seen as enabling important environmen- tal and traffic flow benefits. In addition, V2I technologies are viewed as fundamental to smart mobility applications much more so than in the United States, where a viable business case for V2I is still being discussed. Significant cross-border CV research and development activities underway in Europe include the following: • Following an agreement between the German, Dutch, and Austrian transport ministries, the relevant high- way operators and partners from the automotive industry have launched a cooperative C-ITS corridor from Rotterdam to Frankfurt am Main to Vienna. It will be deployed gradually and enables the exchange of traffic information between vehicles and the road-

A P P E N D I X A : W H I T E P A P E R 39 side infrastructure and information flows among vehicles equipped with cooperative systems (see http:// www.itsinternational.com/categories/networking- communication-systems/features/tri-nation-cooperation- on-c-its-corridor/). • C-Roads is an open platform created by the European Commission and Member States to develop harmo- nized specifications for C-ITS. It was to start develop- ing interoperability validation tests by fall 2017 (see https://www.c-roads.eu/platform.html). • NordicWay is a C-ITS corridor project between Finland, Sweden, Norway, and Denmark. The proj- ect will develop a V-shaped corridor linking Oslo, Gothenburg, Copenhagen, Stockholm, and Helsinki. NordicWay is focused on demonstrating the concept of C-ITS via cellular 3G and 4G/LTE communication, and it will involve about 2,000 equipped vehicles (see Nordicway.net). • UKCITE (UK Connected Intelligent Transport Envi- ronment) is a collaborative project between vehicle manufacturers, communications companies, academia and local authorities to create a 40 miles of urban and inter-urban roads equipped with LTE, ITS-G5 and WiFi to investigate their use in V2X applications to reduce congestion, provide entertainment and deliver improved safety performance (see https://www.cwlep. com/news/uk-cite-project). AV Technologies Safety is a primary motivation for AV development in both the United States and in Europe. As more of the driving task is switched to the automated driving system with SAE Levels 3–5, AVs should mitigate a significant portion of the crash risk stemming from human error. This benefit is cited even subsequent to four known AV fatalities since 2016. Safety (or trust in the technology) has also been cited by several studies as an influencing factor in public acceptance and adoption of AVs (Zmud et al. 2016b, Smith and Anderson 2017, Kolodge 2017, Sener et al. 2018). Unlike the hands-off, market-driven regulatory approach that is the norm in the United States, European countries have taken a much more public-safety-oriented approach. Still, several countries in Europe have welcomed automated vehicle tests on their public roads. Deployers of the technol- ogy need permission, but the procedure may be rela- tively simple, as in Finland: https://www.trafi.fi/en/road/ automated_vehicle_trials. Continuing to test AVs on public roads is critical to development. The machine learning algorithms that govern AV performance currently rely largely on experi- encing various road conditions and situations. Current common belief is that the more miles/kilometers that AVs travel on different roads, in different environments, and under various weather conditions, the more quickly their safety improves and their capability to monitor the surrounding environment to enable observations of other road participants, etc., improves. But since vehicles at SAE Levels 3–5 are not yet on the market, those miles are not accumulating very quickly. Validation methods are ongoing research topics. Evidence on AV performance vis-à-vis a human driver is sparse. One recent study in the United States compared, via simulation, AV crash rates to data from the Strategic Highway Research Program (SHRP) 2 naturalistic driving study (NDS) (Blanco et al. 2016). The research found that self-driving cars in automated mode had significantly fewer crashes than conventional vehicles; however, results were caveated because of the low exposure of self-driving vehicles (about 1.3 million miles in the study) compared to the SHRP 2 NDS (over 34 million miles). In another study, Kalra and Groves (2017) modeled and compared two scenarios: (1) AVs are publicly available for early purchase when slightly safer than human drivers and (2) when market availabil- ity is delayed until AVs are nearly perfect. They found putting vehicles on the road sooner (even if not perfect) can save more lives and improve vehicle performance more quickly than waiting for perfection. Other research has indicated that AVs could address several of the key causes of traffic crashes. • Recognition Errors. For AVs, the impacts on recog- nition errors vary by level of automation. For Level 3, the automated driving system monitors the driving environment and is in control of the driving task. It may request intervention from the human driver at any time, particularly in dangerous situations (e.g., unusual traffic patterns or inclement weather). Much research suggests that this task switching is difficult to do and may exacerbate crash risk (Jannsen and Kenemans 2015; Trimble et al. 2014). At Levels 4 and 5, the automated driving system assumes all aspects of the driving task and does not expect a human driver to intervene. We can expect these vehicles could reduce crashes caused by human recognition errors. But the automated driving system is learning from the driv- ing it experiences as an iterative process. It is basi- cally learning from itself, and so may not know how to behave in unknown situations. In some cases the response may lead to a crash. For example, analyses of accident reports filed by different AV manufacturers testing in California indicated that the most frequent accident was rear-end collisions, happening with a fre- quency that is double that of conventional cars (Favarò et al. 2017). Interestingly, research has indicated that a Level 2 technology (i.e., autonomous emergency

4 0 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S braking (AEB) technology in current model passenger vehicles) led to a 38% overall reduction in rear-end crashes for vehicles fitted with AEB compared to a comparison sample of similar vehicles (Fildes et al. 2015). • Decision Errors. AVs in control of the driving environ- ment (Levels 4–5) that obey traffic laws may reduce decision errors. At Level 3, the driver would remain in control of the driving task and thus, still be in a posi- tion to make decision errors or to disobey or misuse the system. • Performance Errors. At high and full automation (Levels 4, 5), the automated driving system is in con- trol of the driving task and performance errors could be reduced in many situations. However, there is the possibility of overreliance on the automated driving system or driving skill degradation. This is also true for lower levels of automation. This was in fact what the U.S. NTSB found in its investigation into a fatal Tesla crash in May 2016, saying that a probable cause was the driver’s inattentiveness due to overreliance on Autopilot (currently a Level 3 technology) (Bhuiyan 2017). Tesla has since modified Autopilot to warn drivers more frequently to keep their hands on the steering wheel. After three warnings, the system can- not be engaged without stopping and restarting the car. Tesla has also modified how Autopilot’s radar and camera sensors interact to improve the vehicle’s abil- ity to recognize obstacles (Boudette and Vlasik 2017). • Non-Performance Errors. Sleep was the most com- mon critical reason among non-performance errors. A sleeping driver might experience a performance gap in taking over the wheel of a Level 3 AV. When designing for higher levels of automation, a driver should expect to be able to sleep, to enable a high probability of enhanced safety. At Level 3, sleeping would be misuse of the AV. An EU Horizon 2020 project, ADAS&ME, is cur- rently evaluating how the use of C-ITS and automated safety functions, together with unobtrusive driver moni- toring, can compensate for human errors such as those discussed above. AVs might introduce new errors as more of the driving task is switched to the automated driving system; many technologies (i.e., sensors, motion control, trajectory planning, driving strategy, situational awareness, etc.) need to operate effectively so that the vehicle performs at least as well as a human driver (Trimble et al. 2014). New types of vehicle errors could stem from premature release of hardware or software as in the Tesla Autopilot example or inadequately maintained vehicles by owners (private, fleet) or manufacturers. Also, the safe opera- tion of AVs in adverse weather conditions is uncertain (Boston Consulting Group 2015). Snow might cover lane markings so these are not readable by lidar and cameras mounted on vehicles. Snow, frost or ice cov- ering the sensors also causes problems, not only when they cover the lane markings. Heavy rain might damage the lidar mounted on a car’s roof, causing technology failure. While automated driving may introduce new types of driving behaviors and new crash morphologies, it should be the case that the causes of collisions involving auto- mated vehicles can be well characterized through recorded sensor data. Early research suggests that AV technologies have promise in mitigating traffic crashes, but their safety benefits are not guaranteed. Testing of the technologies is necessary for establishing safe operations. In Europe, many new testing activities and demonstration projects at the national and European level are emerging. Examples of the test implementations in Europe include the following: • CoEXist (i.e., AV-Ready’ Transport Models and Road Infrastructure for the Coexistence of Automated and Connected Vehicles) funded under H2020, aims to increase the knowledge of road authorities in transi- tioning toward a shared road network with increasing levels of AVs using the same road network as con- ventional vehicles. The project entails: (1) transport modeling, (2) tool building, and (3) simulated use cases in four road authorities (Gothenburg, Helmond, Milton Keynes, and Stuttgart, see https://www.h2020- coexist.eu/). • Volvo DriveMe Pilot is deploying 100 Volvo XC90s in the first deployment of a Level 3 automated driv- ing system on public roads with non-professional test drivers. The vehicles, equipped with a beta ver- sion of Volvo’s IntelliSafe Autopilot are provided to real-world users for typical commuting and daily use (see https://www.testsitesweden.com/en/projects-1/ driveme). • L3PILOT is a large-scale test started in September 2017. It is unique due to its size (EUR 36 million EU-funding) and is the first in the world to test such a comprehen- sive array of different automated driving functions for passenger cars. L3PILOT involves 34 partners includ- ing 13 Car Manufacturers a large number of systems and component suppliers and leading universities and research institutes. Trials will be carried out in 11 Euro- pean countries, with 100 vehicles and 1000 test drivers. The tested functions cover a wide range from parking to overtaking, and urban intersection driving (see www. l3pilot.ue/index.php?id=26). • AUTOPILOT is a large-scale pilot project started in January 2017 focusing on the autonomous vehicle

A P P E N D I X A : W H I T E P A P E R 41 in a connected environment, enabling the emergence of connected ecosystems supported by open technologies and platforms. The 5GCar started in June 2017 as a large research and innovation project developing the 5G connectivity technologies for automated cars and will evaluate the existing and future spectrum usage for that purpose and contribute to the standardiza- tion efforts in the field (see https://cordis.europa.eu/ project/rcn/206508_en.html). • Truck platooning is the term used to describe trucks using connectivity and automation to follow each other at a very short distance to save fuel and reduce CO2 emissions. The ENSEMBLE project (EUR 20 million EU-funding) will start in summer 2018 and will sup- port the standardization of communication protocols for multi-brand platooning by 2021. The study led by TNO will see collaboration among the six major Euro- pean truck manufacturers (Daf, Daimler, Iveco, MAN, Scania, and Volvo). See http://ec.europa.eu/research/ participants/portal/desktop/en/opportunities/h2020/ topics/art-03-2017.html. • In the next years, more large-scale demonstration pilot projects to test highly automated driving systems for passenger cars, efficient freight transport operations and shared mobility services in urban areas, funded under “Horizon 2020” can be expected. The UK Department for Transport has the stated aim of getting driverless cars on the roads by 2021. Three deployment-pilot projects funded by Innovate UK are currently ongoing. The GATEway project is validating a series of different use cases for AVs, including driverless shuttles and automated urban deliveries on the Green- wich peninsula; UK Autodrive is deploying self-driving pod cars in pedestrian zones in Milton Keynes and Cov- entry; and Venturer is currently moving from simulator studies to applied experiments in real vehicles in con- trolled environments in Bristol and the South Gloucester region (Dennis and Spulber 2017). In the United States as of February 2018, testing of SAVs on public roads is through 17 active pilots in eight states (Stocker and Shaheen Forthcoming). The states are California, Arizona, Washington, Michigan, Pennsylva- nia, Florida, Texas, and Massachusetts by companies such as Waymo, Uber, Easymile, Ford, Navya, Cruise/ GM and Drive.ai. After the fatality caused by an Uber vehicle in Arizona in March 2018, Uber suspended test- ing in North America. The majority of these pilots are targeting Level 4 technology in which a human opera- tor does not need to control the vehicle as long as it is operating in a suitable operational design domain given its capabilities. They are operating as one of two types: (1) on private roads and in planned communities and (2) on public roads and city streets. Shared Mobility As with CAVs, shared mobility operations have the poten- tial to both mitigate and exacerbate human-error caused traffic crashes. In terms of the former, shared mobility operations could mitigate traffic crashes by providing an alternative to driving for some at-risk drivers. Driving under the influence of alcohol (DUI), or impaired driv- ing, is a major contributor to crashes and fatalities on roadways. Proponents argue that ridesourcing services offer a safe transportation option for individuals who have been drinking, particularly among young adults, who are both more frequent users and a segment of the population that may drive while impaired (Elgart et al. 2016). However, research in this area is scarce. While anecdotal evidence suggests that ridesourcing is being used by individuals who go out drinking, formal research lacks data to attribute reductions in impaired driving and improved safety to any one factor, such as ridesourcing services (Shirgaokar 2016). In terms of exacerbating crash risk, increased con- gestion at the curbside not only increases the potential for vehicle crashes but also for crashes with other road users (Rogers 2017). The “curb” is home to bikesharing programs, cycling lanes, ridesourcing passenger pick up and drop off, and goods delivery. Some cities have also set aside curbside space for carsharing services. As such, curb management for congestion and safety has become a priority for many cities. We should note than in addition to congestion and safety issues, there are equity issues pertaining to the use of curbside space for a private business or non-profit purpose, as well as for competing operators and modes (Shaheen et al. 2016). As of yet, curbside management for congestion caused by ridesourcing or increased goods delivery operations does not appear to be as widespread an issue in Europe as it is in the United States. Personal security concerns have been raised about many innovative mobility services, as they have been historically for the conventional for-hire industry (Trans- portation Research Board 2016). Incidents involving safety of passengers receive intense media attention, although little research has actually documented the prevalence. However, ridesourcing technologies (and similar apps being adopted by the taxi industry) may mitigate risks to passengers and drivers by documenting the details of trips and removing anonymity, as may the cashless transactions made possible through ridesourc- ing or mobility-on-demand billing systems. In the United States, many local authorities and municipal, regional, and state governments are reviewing public safety regu- lations for ridesourcing and other shared mobility ser- vices. For example, much attention has been given to the inconsistencies between the background checks applied to taxi versus ridesourcing drivers and of different vehi- cle inspection requirements for the two types of services.

4 2 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S Trust in AV technology is a barrier to acceptance and use. So a question of interest is what is the influence of the absence of a driver in autonomous ride-hailing vehicles? There is sparse research on the topic. A Kelly Bluebook survey (2016) found that respondents pre- ferred using ridesourcing services with human drivers to using them as self-driving vehicles (respectively, 56% vs. 44%). But among current ridesourcing users, there was a preference for using them as self-driving vehicles (51% vs. 49%). That study prompted another study that tested the hypothesis that current ride-hailing users will be early adopters of automated vehicles (Sener et al. 2018). This latter study also found that current ride- sourcing users would be almost twice as likely to accept and use automated vehicles as non-users. Will there be many single occupancy trips or more high occupancy trips? With these services, a driver may pick up more than one rider going in the same direc- tion. Most frequently cited reasons for this were incon- venience and discomfort associated with riding with strangers, especially in the absence of a designated driver. Why Cybersecurity Is an Important Socioeconomic Impact Issue for CAVSM Cybersecurity—in the context of vehicle systems, refers to security protections for systems in the vehicle that actively communicate with other systems or other vehi- cles (Bryans et al. 2017). While cybersecurity issues are a challenge for CVs, security becomes a bigger concern with Level 4 and Level 5 AV vehicles, in which software and connectivity play a much bigger and more critical role for the safe driving of vehicles. Unlike traditional vehicles, AVs may be vulnerable to cyber-attacks that can spread from vehicle to vehicle, which may constitute a new type of safety threat. In the case of a cyber-attack the safety of passengers in an AV and other road users could be at risk. In a case of hacking and stopping a fleet of AVs, the transportation system could be halted with potential safety reduction (even though no real case of malicious car hacking has been reported yet). Miller and Valasek (2015) exposed the security vulner- abilities in automobiles by unmaliciously hacking into cars remotely, controlling the cars’ various controls from the radio volume to the brakes. All entry points into the vehicle, such as Wi-Fi, the OBD-II port, and other points of potential access to vehicle electronics, could be potentially vulnerable to real-time intrusion (hacking) that could affect the mechanical operation of the vehicle. A large number of vehicles communicating to/with each other is essentially an ad hoc, self-forming network of devices with no server-side security (McCormick 2017). Cybersecurity, therefore, is a new factor that shapes the existing crash externality. Since a very small percentage of accidents are caused by mechanical errors, this should have little actual negative consequences in terms of the safety benefits of CV or AV technologies, as the $1.2 bil- lion Toyota settlement, after a four-year criminal probe into its handling of a spate of sudden accelerations in its vehicles, highlighted. However, one major high-profile mechanical failure of an AV could have profound impli- cations for technology deployment. Security by Design, Standards, and Legislation The U.S. DOT has adopted a “security by design” principle as it develops the system architecture for con- nected vehicles–meaning that cybersecurity systems will be built in from the start. When people speak of secu- rity by design, they often refer to a broad spectrum of activities and approaches used to build stronger secu- rity. “Spectrum” is an accurate term for this concept, as it spans lifecycle activities and functional domains, i.e., consideration of requirements, definitions, design, development, testing, and maintenance. Cyber solutions need to be developed in the context of security vehicles not just adopted from other industry sectors (Kitayama et al. 2014). While the end-to-end security design prob- lems for the IT industry have been developed, these are not necessarily applicable to vehicles. There are signifi- cant differences between securing IT equipment (such as servers and PCs) and securing vehicles. One of the main differences is that with vehicles human safety is a pri- mary design consideration. In addition, the lifecycle of a vehicle is often much longer than the lifecycle of many PCs and related IT equipment. While the U.S. DOT promotes the security by design concept, there is only voluntary best practices guidance on vehicle cybersecurity. There are also industry stan- dards being developed through the SAE. These include SAE J3101–Hardware protected security for ground vehicle applications and SAEJ3061–Cybersecurity guidebook for cyber-physical vehicle systems. On the other hand, in Europe there is EU-wide legisla- tion on cybersecurity (Directive on Security of Network and Information Systems) that was adopted by the Euro- pean Parliament in August 2016. Member States were required to transpose the directive into their national laws (https://eur-lex.europa.eu/legal-content/EN/TXT/ ?uri=CELEX%3A52017XX0720%2801%29). Among other measures, Member States are required to set up a Computer Security Incident Response Team and a competent national NIS authority to promote swift and effective operational cooperation on specific cyber- security incidents and sharing information about risks. Businesses in sectors that are identified by the Member

A P P E N D I X A : W H I T E P A P E R 43 States as operators of essential services have to take appro- priate security measures and notify relevant national authorities of serious incidents. Also key digital service providers (search engines, cloud computing services and online marketplaces) have to comply with the security and notification requirements under the new directive. economIcS and Workforce ISSueS CAVSM on economics and the workforce in the United States and the European Union. Why Economics and Workforce Are Important Socioeconomic Impact Issues for CAVSM Whether it is the movement of goods from manufac- turer to marketplace, material to the manufacturer, or the movement of employees from home to office, mobil- ity is directly associated with economic prosperity (e.g., Eddington 2006). The introduction of CAVSM services could influence the availability, cost, and efficiency of mobility services with an associated impact on local, regional, and national prosperity. The ways in which CAVSM services are deployed and operated may have differential impacts on how the benefits of automation are distributed. In addition, the presence or absence of connectivity may shape where CAVSM services can be deployed suc- cessfully. The use of robots is often linked to improving efficiency by tackling tasks that exhibit one or more of the three ‘D’ characteristics: dull, dirty, and dangerous (Murphy 2000). Automation of driving can therefore increase the efficiency of transport by providing safer, more reliable transportation. However, task automation is typically associated with a reduction in the number of employees and/or the training required to deliver that task. This is especially true when an employee repre- sents a significant element of the operating costs for that system. For a taxi in Zurich, a driver is estimated to rep- resent 88% of the operating costs of the vehicle (Bösch et al. 2017); for a bus in Zurich, the driver is estimated to represent 55% of the operating costs of the vehicle (Op. cit.); while the average marginal costs for a truck driver in the United States are estimated to represent 43% of the operating costs of the vehicle (Hooper and Murray 2017). Consequently, there are significant eco- nomic efficiencies to be achieved if an automated system could replace the driver and for use cases where there is economic benefit, the switch may happen rapidly. How- ever, it is over-simplistic to assume that all costs associated with the driver would be saved by introducing automated vehicles. The purchase/leasing and maintenance of these vehicles would represent a greater expense than tradi- tional vehicles while other challenges to business models may emerge as roads authorities learn how to manage the deployment of automated vehicles effectively. A typical car spends the majority of its life static, with use at less than 5%. The ability to share vehicles, either by having a single vehicle serve multiple individual custom- ers sequentially or by pooling individuals taking similar journeys into a single vehicle, could increase vehicle use and thereby unlock previously unattainable efficiency Key Takeaways • Access to opportunities, underpinned by mobility, is a key enabler of prosperity that can be enhanced by new services facilitated by connectivity and automation. • Driven by the potential for operational efficiency, organizations are exploring opportunities for CVs and AVs in long haul trucking operations and urban deliveries. • A proliferation of connected, automated (as trials) and/or shared passenger trans- port services have emerged, particularly in cities in both the United States and the European Union. • This could have a range of impacts includ- ing changes to mode choice, acceptability of trip length, land use values, accessibil- ity of employment, retail and congestion impacts. • While new jobs may be created in the operation and development of CV and AV services, there is a high likelihood of job losses from driving-focused roles. • Given these societal impacts, it will be important that regulatory authorities are aware of these impacts and can act to maximize the benefit and minimize harm from the proliferation of CAV services. The development and deployment of CAVSM is underpinned by envisaged safety and efficiency benefits. However, CAVSM will not flourish without also deliv- ering sound economic returns. Innovation in CAVSM technologies is also creating opportunities for new forms of transportation to emerge. These have the potential to alter significantly the number and types of jobs associ- ated with the movement of people and goods. This sec- tion therefore explores the socioeconomic impacts of

4 4 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S benefits. The prerequisite is that these services are attrac- tive and accessible enough to be used. Shared vehicle ser- vices are made feasible by connectivity (enabling users to find rides with others traveling to the same or nearby destinations) and can potentially be made more efficient by automation (by reducing the operational costs and increasing operational flexibility of the shared vehicles) (Greenblatt and Shaheen 2015). Freight Transport The movement of goods is often cited as a low-margin activity (e.g., Caballini et al. 2017). As a result, improve- ments in efficiency are aggressively pursued with the freight sector pushing innovations in connected vehicle and goods tracking technologies. Automation of the driv- ing task is attractive for the industry to reduce the eco- nomic cost and physiological constraints (e.g., fatigue) on freight operations. However, the activities of a goods driver can extend beyond the task of driving (including vehicle checks, load checks, administration etc.) and not all freight delivery tasks entail the same driving complex- ity. These factors may therefore guide the emergence of automated vehicles for the delivery of goods, with the early opportunities likely to appear in highly controlled environments, such as ports, airports and mines, where the complexity of the automation task is reduced and the goods being moved are well organized. Indeed the Port of Rotterdam has operated fully automated vehicles for container movements since 1993 (Bishop 2000) and fully automated mining trucks are now well established (Simonite 2016). In a similar manner, automation of freight vehicles on the public road is likely to emerge in locations where there is control over the environment and where the economic returns are greatest. This has seen a range of companies promising automated high- way driving for trucks (e.g., Uber, Waymo, Embark, and TuSimple). Otto (the former name for Uber’s automated trucking initiative) demonstrated delivery of a ship- ment of beer along I-25, Colorado, in partnership with Anheuser-Busch. The articulated truck apparently drove 120 miles from on-ramp to off-ramp with no intervention from the human driver present (Fitzpatrick, 2016). Based on the eventual introduction of such technology, a PWC report estimated automated, long-haul trucking could reduce total transportation costs by nearly 30% through 2040, assuming aggressive adoption of automated truck- ing (PWC & MI Manufacturing Institute 2018). In Europe, real-world trials of truck platooning have taken place with vehicles from manufacturers such as Daimler, Volvo, Scania, and DAF participating. For exam- ple, convoys of trucks from each manufacturer completed journeys from different parts of Europe, converging on the Dutch port of Maasvlakte. Drivers were present in all trucks but only the lead vehicle was fully driven by a human driver. ElectronicE connections between the lead truck and following trucks managed acceleration and braking to enable closer following distances. Such dem- onstrations have shown real-world improvements in fuel efficiency of 8% (Chan et al. 2012). In an industry where fuel costs represent an average of 21% of truck operating costs in the United States (Hooper and Murray 2017) and 26% of operating costs in Europe (Meszler et al. 2018), this represents a significant potential improvement in profitability, if platooning can occur on significant por- tions of journeys and if the technology to deliver platoon- ing is proven as safe and is not prohibitively expensive. Furthermore, to date, trials of platooning have tended to be between vehicles from the same manufacturer. To maximize the opportunity for platooning to take place, it will be necessary to achieve multi-brand platooning where trucks from different manufacturers can platoon interchangeably, such as with the ENSEMBLE project mentioned in the previous chapter. The introduction of automated vehicles to the freight industry has caused concerns about job losses: the role of human truck drivers will be taken by automation technology (Beede et al. 2017). The International Trans- port Forum (ITF) has predicted that up to 4.4 million of the 6.4 million professional trucking jobs in the United States and Europe could be eliminated by autonomous technology (ITF 2017). This concern is underlined by the popularity of truck driving as a form of employment; a recent study (Bui 2015) indicated that truck driving was the most common job title in the majority of U.S. states. However, in both the United States (Costello 2017) and Europe (e.g., Todd and Waters 2018), the logistics industry has seen a shortage of drivers. In the short term, it may be that automated vehicles mitigate this human driver scarcity. In the longer term, when automation may play a greater role in the movement of goods by road, the transition away from truck driving as a common form of employment may be effectively man- aged. This may include roles in managing operations from regional control centers and a range of different tasks associated with the maintenance and management of automated delivery vehicles. There are also benefits from automation that do not result in the loss of delivery drivers from the workforce. Truck drivers in the United States (FMCSA 2011), Euro- pean Union (European Commission 2006) and globally (e.g., Australia: National Transport Commission 2006) are strictly regulated in the interests of safety and fair working conditions. If automation can be proven to man- age long periods of highway driving safely and efficiently and if the workplace environment for the driver can be made acceptable (toilet facilities, refreshments, connec- tivity etc.), it may be possible to lengthen the operating window for truck operations leading to increased delivery

A P P E N D I X A : W H I T E P A P E R 45 efficiency. While the AV manages driving for long stretches of the highway, the human driver can engage in other administrative tasks or relax. The driver would need to be able to resume controls when needed through a managed and practiced procedure and with appropriate failsafe systems, should a driver fail to respond for any reason. These changes in the responsibilities of the role of the driver may require a different skillset and additional or different training to maximize productivity. The safety and economic benefits of driver assistance technologies have driven their adoption in regulation in the European Union with lane departure warning and automatic emergency braking systems made mandatory from 2014 (European Commission 2009). In the United States, there has been significant trialing and develop- ment of platooning technology to improve vehicle fuel efficiency (e.g., Peloton, see Simpson 2018). If the imple- mentation of higher levels of vehicle automation can fur- ther reduce the incidence of collisions, these benefits can be extended still further with greater vehicle uptime and reduced insurance and repair costs, but also with the employment implications discussed above. The ability for vehicles to move without human opera- tors means that vehicles can be developed to suit the deliv- ery requirement and transport environment without the need to consider accommodation of a driver. This has led to a proliferation of small, robotic vehicles (e.g., Starship Technologies, Nuro) intended for very low-cost, short- range urban deliveries. Such vehicles may help tackle the impact that the growth of online deliveries has had on city traffic (Visser et al. 2014). The Starship vehicles have undertaken trials of grocery (Karasin 2017) and food (Gerrard 2017) deliveries. If this is proven to work success- fully, this could enable a more significant transformation toward the sharing economy, where material owner- ship of items is less critical if they can be delivered and returned at very low cost and at user convenience. How- ever, there has been some resistance to their deployment (Wong 2017). There could be serious induced demand and VMT/GHG effects, which could also limit accep- tance. As with Jevons’ paradox (Jevons 1865), an unintended consequence of increased delivery efficiency might be a dramatic increase in the number of deliveries being undertaken. While trials of small numbers of vehicles may be seen as acceptable, regulation may be needed to mitigate the effect of their presence on the experiences of pedestrians, cyclists, and the wider traffic environment when deployed in larger numbers. Passenger Transport Many current automated vehicle trials are offering differ- ent varieties of passenger transport. These include Waymo (Korosec 2018) and Aptiv/Lyft (Etherington 2018), trial- ing automated passenger car services; Navya (Christie et al. 2016) and Easymile (Robarts 2015), offering automated bus operations with up to 12 passengers per vehicle; and Aurrigo (Parmenter 2017) and nuTonomy (Ackerman 2016), seeking to deploy small, personal vehi- cles dedicated to urban environments (discussed in prior chapter, referencing Stocker and Shaheen Forthcoming). Although they are exploring different business models, each is focused on the opportunity for using automated vehicles to move passengers in towns and cities (Stocker and Shaheen 2016; Stocker and Shaheen 2017). These approaches stem from a belief that private cars contrib- ute to urban congestion, and city mobility can be signifi- cantly improved by connectivity and automation. While the role of AVs may enhance transportation opportunities in urban environments, it should also be recognized that some European (e.g., Amsterdam, Copenhagen) and U.S. cities (e.g., Atlanta, Chicago) place a significant empha- sis on the role of active travel (walking and cycling) for cities. This is captured by the London Mayor’s Trans- port Strategy (Greater London Authority 2018), while a vision for the balance between existing modes, AVs, and active travel is neatly described in the ‘Blueprint for Autonomous Urbanism’ (NACTO 2017) produced by NACTO. Each document sets out how technology and urban design should be used to support the needs of the city by applying people-centric design. The focus on urban environments for AVs is logical given the density of customers and need for new forms of mobility in those areas. However, (re-)connecting rural areas may also provide opportunities for CAVSM. Profit- ability is challenging when confronted with practicalities of operating large buses to enable transport outside of cities. However, connectivity and shared transportation can be used to enable greater demand responsiveness, while automation may be able to reduce operating costs further (see for example, http://innovativemobility.org/ wp-content/uploads/Mobility-on-Demand-Operational- Concept-Report-2017.pdf). This may support mobility for older travelers who are more likely to live in rural areas and may have to give up driving, having been car dependent. Similarly, such transportation may help younger residents access educational facilities and broaden the employment horizons for rural residents (Shergold et al. 2016). Connectivity in the workplace has threatened to revo- lutionize travel for many years. Nearly 40 years ago, it was speculated that commuting would fall dramatically as connected workers would be able to log-in from home (or wherever) to accomplish their office duties (Toffler and Alvin 1980). However, although remote working is now possible for many office-based employees, demand for mobility has not (yet) diminished as predicted; com- muting by road remains a significant component of the working lives of U.S. and EU citizens. However, urban

4 6 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S areas commonly tend to exhibit a traveling time-distance radius (or travel time budget) of 45 minutes (Muller 1995). The potential to use CAVSM for commuting could dramatically change lifestyles and working practices and change this time-distance radius. First, by increasing the speed and efficiency of transportation, the bound- ary of the time-distance radius may be extended out- wards away from the center. Second, the comfort and convenience afforded by AVs may enable greater flex- ibility in the use of time when traveling, for example, enabling work on the move, such that commuters may be more comfortable traveling greater distances (Maia and Meyboom 2018). Also traveling during peak hours may increase if delays are perceived to cause less harm. This may result in induced demand and increases in vehicle miles traveled and emissions if demand is not managed by pricing regimes or capacity is not increased in line with the increased demand. These secondary effects may mitigate the potential societal benefits of AV transporta- tion. A later section of this chapter presents a discussion of congestion impacts. Shared transportation services are enjoying significant growth, enabled by connectivity and by the market pen- etration of smartphones, through which such services can be accessed (Shaheen et al. 2016). Other societal trends support this growth. In particular, the tendency for millennials to drive less than older age groups and being more likely to use technology to substitute for travel (Op. cit.). A variety of new concepts have emerged including carsharing, ridesourcing/e-hailing, microtran- sit, and digital ridematching (Op. cit.). There are many potential impacts associated with the uptake of such ser- vices. For example, in addition to potential reductions in congestion and pollutant emissions, joining a carsharing scheme is estimated to save U.S. households $154–$435 per member per month (Shaheen et al. 2015). In areas that are poorly served by public transit sys- tems (and where their introduction might be expensive or impractical), CAVSM services may provide vital accessibility, helping local residents and businesses meet their mobility objectives. This includes helping citizens to achieve educational, medical, social, and employ- ment objectives and businesses to gain access to more efficient supply chains and a broader market. This in turn may help the local authorities by supporting citi- zens into employment and reducing dependence on pub- licly funded services for transportation and medical care. However, it should be noted that improving connectivity to a region would be likely to cause the value of proper- ties within that region increase. By reducing affordability of properties, there is a risk that low-income residents would be pushed further away from city centers, exacer- bating their travel challenges. As society becomes increas- ingly urbanized, CAVSM services may help the viability of rural communities through improved transportation (Meyer et al. 2017). Research has also indicated that the use of ridesourc- ing services has led to reduced use of public transpor- tation and increased congestion in some urban areas (Feigon and Murphy 2016; Rayle et al. 2016; Clewlow and Mishra 2017; Schaller 2017; Gehrke et al. 2018; Hampshire et al. 2017). Because of this, vehicle occu- pancy rates in shared automated vehicles are important. Both Lyft and Uber offer shared-ride services, known as uberPOOL and Lyft Line, which are cheaper than regu- lar ridesourcing services but can be more expensive than public transit. Sener et al. (2018) found that likely users of automated ridesourcing services are much less likely to use the pooled version (39%) than the single occu- pancy version (61%). The addition of automation into the road transport sector need not (and should not) mean that all public transit vehicles operate with no onboard employees. For example, in providing bus services for the elderly or disabled, the role of the driver goes beyond vehicle operation. A driver provides social contact, sup- port in accessing the vehicle and information about the trip (SCR 2018). Freed from the responsibilities of driv- ing, an operative on board such a vehicle would have more capacity to perform these ‘added value’ duties, thereby delivering a better overall service to users. Job Market Impacts One of the biggest impacts of connectivity on mobil- ity is the growth of ridesourcing services, such as Uber and Lyft. The development of ridesourcing has created new ways in which people can monetize car ownership and driving ability. Flexibility in where, when and how one chooses to offer service through one or more ride- sourcing services means that drivers can choose how much of their time they wish to spend working in their service. Without supervision from a full-time employer, ridesourcing activity is only possible through continuous connectivity that is required to manage operations to find customers and to secure payment (along with sup- port for routing, traffic avoidance etc.). This style of working is an example of the so-called ‘platform’ or ‘gig’ economy, in which workers operate as freelancers, choosing their level of availability for work and where payment is not in the form of a regular salary but is determined by the number of customers and types of job they are able to complete. This style of working offers useful flexibility for drivers and has created convenience for riders but is associated with a distinct lack of social protection for platform workers (Forde et al. 2017). It is a model that is also threatened by the emergence of automated vehicle services. Uber and Lyft are investing heavily in the exploration of such services to ensure they can take advantage of this tech- nology and compete with emerging rival providers of

A P P E N D I X A : W H I T E P A P E R 47 automated vehicle services such as Waymo, Ford, GM and Daimler/Bosch. This significantly changes the style of operations for a ridesourcing service, shifting more toward the maintenance and management of an AV fleet rather than relying on driver/owners to maintain vehi- cles appropriate for taxi operations. While the deploy- ment of automated vehicle taxi services is likely to occur incrementally; starting in small regions of cities, once proven to be successful (safe, well received, profitable), their growth could be rapid, reducing the size of this gig economy for on-demand drivers. It is important to rec- ognize the transitional state between current ridesourc- ing services and full SAVs. There may remain a market (and a need) for human-driven vehicles where by choice or by operational requirement (e.g., a VIP whose safety cannot be entrusted to a vehicle that always stops for obstacles or can potentially be hacked) a human opera- tor is preferred. At the same time, the potential for AVs to reduce employment cannot be ignored. A U.S. study in 2017 identified that 81% of U.S. adults anticipated that many people who drive for a living would lose their jobs (Smith and Anderson 2017). In 2015, there were 15.5 million workers in the United States employed in roles that could be affected by the introduction of auto- mated vehicles, representing one in nine of the available workforce (Beede et al. 2017). In this report, the authors distinguish ‘motor vehicle operators’ (those for whom driving to transport people or goods is a primary occu- pation (e.g., truck drivers, bus drivers, taxi drivers) and ‘other on-the-job drivers’ (those who drive to deliver services or trades (e.g., construction workers, real estate agents, police patrol officers). The authors assert that automation is likely to have a more significant negative impact on members of the former category, who may find it difficult to find alternative work whereas workers in the latter category depend on a range of other skills in the scope of their employment and may therefore adapt more successfully to the introduction of AVs. Since members of the ‘motor vehicle operators’ were likely to be male (88%), not educated to degree level (92%) and less likely to have a health plan or a pension or to live in a metropolitan area, it can be seen that supporting this group through the transition to the world of automated vehicles will be important. Although the report offers a rather negative view on the impact of AVs on employ- ment prospects, it fails to recognize the potential for jobs to be created by their deployment. Three areas stand out: (1) software (supporting the programming and develop- ment of AV platforms); (2) maintenance of automated vehicle fleets (ensuring vehicles are clean and that sen- sors, actuators, other associated systems are operating within acceptable tolerances for automated operation etc.); and (3) data analysts (managing and analyzing the terabytes of data that automated vehicles will produce). That said, it is unlikely that those in roles replaced by automated systems will have been adequately trained by current educational systems to achieve the necessary skillsets required to enter any of these new roles. Goos and Manning (2007) recognized technology as causing a polarization of the job market. There is an increase in demand for well-paid, skilled jobs that involve non- routine cognitive skills and an increase in demand for low paid, low skill jobs that involve non-routine manual skills. Frey and Osborne (2013) highlighted that machine learning and mobile robotics are significantly increasing the number and types of task that can be automated, thereby exacerbating this polarization and with the risk of job automation disproportionately affecting low skill/ low wage occupations. A sanguine examination of the risk of automation for jobs in OECD countries (Arntz et al. 2016) estimated that 9% of all jobs were potentially automatable, somewhat lower than the estimates of Frey and Osborne. The researchers emphasized that it was often the task-related content of occupations rather than the occupations themselves that would be automated. However, for jobs where employees are paid solely for driving (as per Beede’s ‘motor vehicle operators’), it must be assumed that automation will obviate the need for those employees. This specific risk for unemployment in relation to driving must be a consideration for transport authorities in the rush to automation. There are potential employment impacts other than job gains or losses. Ubiquitous connectivity also changes employees’ value of time. Many workers engaged in office-based tasks can remain in contact with colleagues and customers provided that they have telephone and data connectivity. In addition, if transport options pro- vide conditions conducive to work (e.g., seating, refresh- ments, toilet facilities), travelers can gain value from time spent in public transit. Also sufficient privacy is needed if you plan to work while traveling. While this may help workers to achieve a desired work/life balance, it may also affect willingness to travel and acceptance of longer commutes since time spent in journeys can still be use- ful. Applied to CAVs, this could increase vehicle miles traveled since passengers may accept more frequent and longer road journeys during which they can remain pro- ductive. However, the dynamic characteristics of road transport (lateral and longitudinal accelerations) and behaviors associated with computer work (continuously looking at text on screens) may cause some users to feel discomfort associated with carsickness (Diels and Bos 2016) and so assuming that all travel time can be pro- ductive is likely to be false. Driving for work is typically the riskiest activity in which office workers engage (Broughton et al. 2003). As a result, the crash safety ratings (e.g., NHTSA/NCAP, EuroNCAP) of vehicles have become a factor in pur- chasing decisions for fleet managers. Over time, collision

4 8 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S data will establish the relative risk of collision for AVs compared to human-driven vehicles. If AVs are proven to be statistically safer, fleet managers may be obliged to require employees to use automated vehicle services rather than drive for themselves, not just for reasons of productivity but under their duty of care responsibilities. Congestion Impacts Road congestion has a significant impact on economies (Goodwin 2004; Hartgen et al. 2009). It is experienced when demand for a road exceeds capacity and results in delay. For passengers, delay can mean the loss of pro- ductive time, missed meetings or missed onward travel connections; while for freight, delay might mean finan- cial penalties for the supplier, loss of business or spoiled goods. The impact of CAVSM on congestion is depen- dent on many factors. The first factor is the type of vehi- cle used. If all human-driven vehicles were exchanged for automated vehicles, (assuming there were significant dif- ferences in the way the vehicles were driven: significantly shorter time headways, smoother interactions at junc- tions) this would have a limited impact on congestion. However, if shared and pooling of vehicles is enabled by connectivity and automation, significant efficiency benefits may be achieved. It has been shown that one shared, automated vehicle can replace nine conventional vehicles but also generate increased vehicle miles trav- eled due to empty operations and repositioning (Fagnant et al. 2015; Greenblatt and Shaheen 2015). However, as TNC operations become more mature, some behavioral adaptations are starting to emerge. There are circum- stances in which such services may increase congestion (Schaller 2017) and attract travelers away from bus (6% reduction) and light rail (3% reduction) services (Clew- low and Mishra 2017). When automated transport facil- ities become available, transport authorities will need to be mindful of these adaptations in approving such operations for service. Although it has been calculated that operating a fleet of shared automated taxis to ser- vice mobility needs within a city may be achievable at drastically lower cost than by private vehicles (see Bösch et al. 2017 for review), it may be cities will need to apply fees to such operations to help manage congestion and access to mobility. The growth of online shopping has changed con- sumer habits. Rather than making infrequent trips to retail zones for multiple products, a shopper can make frequent online purchases, benefitting from rapid, low- cost deliveries of individual (or multiple) items. This has changed the nature of the logistical challenge of delivery fulfilment with the consequence that cities have growing numbers of small goods vehicles making multiple deliv- ery drops. This has an associated impact on traffic and congestion (Visser et al. 2014). At present, the majority of vehicles used for such goods deliveries are vans with combustion engines. A growth in the use of such vehicles is therefore also a source of congestion and emissions (Russo and Comi 2012). However, with growing emphasis on air qual- ity and with the emergence of new technologies, other options are emerging. Examples include GNewt Cargo, based in London, which exclusively uses electric deliv- ery vehicles for duties in the city; and PedalMe, also in London, which serves passengers and deliveries within a five-mile radius of the city center using a fleet of dedi- cated cargo e-bikes. Automation brings the potential for new vehicle formats. Examples include Starship Tech- nologies and Dispatch Robotics, which fulfil local deliv- eries using mobile robots that traverse pedestrian spaces and can carry small packages to a chosen destination. To date, these systems have tended to be used for deliveries of items such as food or laundry. However, their applica- tion could perhaps have further reaching implications. If small items can be delivered at very low cost and on very short notice, the need for ownership of items is reduced. For example, a user could hire a specific power tool for a particular job on a given day, this would be delivered by a mobile robot and then the robot would return to collect the tool at the end of the day. There would be no need for the user to own the power tool, which might otherwise lie unused in storage for the rest of the year. If such schemes became commercially viable, they could transform ownership and usage models for a range of different products and services and empower low-income communities to access these products and services in ways that were previously impractical. However, it should be noted that the deployment of such mobile robots has not been without resistance. In product development, when only a few such sys- tems are being trialed, the presence of the devices in pedestrian areas creates minimal disruption and can be seen as a novelty. There are concerns that in com- mercial deployment, the presence of large numbers of such vehicles could intrude upon pedestrian spaces. In 2017, San Francisco enacted regulations to control the operation of such vehicles, restricting their use to spe- cific industrial regions and sidewalks that are at least six feet wide and requiring the vehicles to be accompa- nied by a human operative (Wong 2017). A further opportunity may be enabled by very low- cost delivery services achieved by CAVs: if a shopper does not need to transport their goods home with them but can rely on delivery of products to a place and at a time of their choosing, they may be freer to choose which transport mode they prefer (see Shaheen et al. 2017). For example, a person buying a bulky or heavy item may not be comfortable cycling home with that item but if low- cost home delivery can be arranged (mediated by CAVs), cycling to and from the store may be seen as acceptable.

A P P E N D I X A : W H I T E P A P E R 49 Impacts on Land Use Values While uncertain, CAVSM has the potential to signifi- cantly impact land use values. For example, the shift toward the convenience of online shopping and the easy parking in out of town shopping malls presents a chal- lenge to city center retail zones (Jones and Livingstone 2018). Higher rents in premium city center locations combined with falling numbers of shoppers means that it can be difficult for such stores to maintain profitabil- ity. However, the emergence of AVs may reverse this decline. Such vehicles would be able to deliver shoppers directly to the city center without worrying about park- ing. Furthermore, the vehicle would manage the stresses of having to drive in city center traffic while occupants could be shown advertising messages promoting prod- ucts and services on offer in the city center. If CAVSM were to restore the popularity of city centers as destina- tions, it could provide the impetus for wider rejuvena- tion of such districts. In fact, the emergence of highly automated vehicles challenges the need for parking infrastructure in cities and reduces the need for housing developments to include parking, which have significant economic impli- cations (reviewed in Litman 2016). A significant area of land in urban regions is dedicated to vehicle parking. An analysis of selected areas within Sacramento found that parking covers about 11% of the downtown, 26–39% of the industrial areas studied, 30–57% of commercial areas analyzed and 6–26% of the residential areas exam- ined (Chester et al. 2015). The availability of parking is an influential factor in trip planning and even choice of home location. V2I connectivity is enabling new park- ing services to emerge that seek to provide drivers with knowledge of available parking spaces near their desired destination and even to enable booking of and payment for parking slots. Choosing to share vehicles reduces the per-traveler parking demand and parking costs can be shared. The benefits of CAVs are expanded if travel is by shared autonomous vehicles. Zhang and colleagues (2015) used a simulation approach to estimate that the used of shared autonomous vehicles in an urban envi- ronment could reduce parking demand by as much as 90%. If such services reduce the need for parking, the value of land may be significantly affected and existing parking lots can be reclaimed for other purposes. The land hosting existing car parks in city centers may rep- resent a hugely valuable asset that can be repurposed for other means. Similarly, city center shops and amenities that do not have access to good parking facilities may become more valuable if users are able to access stores more effectively by using automated vehicles that can drop off consumers close by. However, while it is possi- ble that AVs will reduce the need for parking, it does not of course eradicate the need completely. AVs will need to be stored, cleaned and maintained in secure premises, the location of which will be optimized to support the opera- tion of the vehicles. Further, city authorities are unlikely to welcome the prospect of free-floating, unoccupied vehicles circulating in traffic while they await assignment to their next journey. Consequently, they will need to find suitable waiting areas. equIty Key Takeaways • There are significant potential socioeco- nomic benefits of CAVs, the distribution of which may change social equity in a posi- tive or negative direction. • CAVs could enhance equity by improving access to opportunities through increased mobility, particularly for those who have fewer options at present. • Potential for inequity in safety and air quality impacts must be considered. • Technology can act as a barrier to new mobility services for those less familiar with its use and/or not able to use elec- tronic financial services. • Integration of private CAV services with public transportation must be carefully overseen to ensure equitable distribution of transport options. Why Equity Is an Important Socioeconomic Impact Issue for CAVSM Social equity for governmental organizations relates to the fair distribution of services across its potential recipi- ents. The subject of equity is an important socioeconomic consideration for CAVSM because it is anticipated that CAVSM technologies may dramatically enhance access to mobility for people and businesses (Lazarus et al. 2018). Access to opportunities (Martens 2012), often mediated by mobility, is a key determinant of prosperity and well-being in society (Eddington 2006). The vehicle designs and technologies used, the markets addressed by CAVSM and the regulations imposed upon such services are all factors that may exert a strong influence over how these mobility enhancements are distributed. For the purposes of this paper and in relation to CAVSM,

5 0 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S we consider equity in terms of the distribution of socio- economic outcomes associated with CAVSM that serve to reduce inequality, deliver a fairer and more just soci- ety and that provide more options (in mobility and in life) to those that have fewer at present. Safety Considerations As discussed in a previous chapter, safety is a potential benefit of connected and automated vehicles. To date, new luxury vehicles have tended to be the vehicles with the best safety systems. Over time, driven by regula- tion, by market demand and/or by economies of scale in deployment of such systems, these safety features should be integrated into less expensive mass-market models. This cascading of safety systems from luxury to mass- market vehicles gradually broadens the safety benefit that such technologies can achieve. However, the drivers that present the highest crash risk on the roads (typically younger, typically male; see e.g., Helman et al. 2010) tend not to drive newer vehicles that are equipped with the latest safety technologies. With the development of vehicle automation, some authors (e.g., Fagnant and Kockelman 2018; Greenblatt and Shaheen 2015) have envisioned a new era of carsharing where users do not own vehicles but access appropriate vehicles on an as needed basis. For sound commercial reasons, the operators of such vehicle fleets will be able to make evidence-based judgments over the vehicles they deploy and the safety systems that are present on each vehicle. Similarly, consumers may be able to select their preferred vehicle supplier based on known safety risks. Furthermore, given greater utiliza- tion of the vehicles within their fleets (and like hire car companies today), operators are unlikely to hold vehicles for more than two years. As a result, a customer of a shared vehicle fleet is likely to encounter vehicles that are relatively new and fitted with the latest safety systems. If such fleet operations are accessible to younger drivers, this may reduce inequity between younger and older drivers in terms of their access to vehicles that are better at avoiding (or mitigating the impact of) collisions. The significant and hard won improvements in road safety achieved in Europe and North America since the 1960s are often attributed to the “three Es” (e.g., Pease and Preston 1967): Engineering; Enforcement; and Edu- cation (sometimes extended to four by the addition of emergency care). A possibility that may arise when con- sidering the future impact of CAVs on road safety is that practicable short-term gains in these traditional areas may be overlooked in favor of supporting the promises of road safety that high levels of vehicle automation are said to offer (Bajaj 2018). It is therefore important in socioeconomic terms to recognize the danger in over- estimating the future benefit of as yet unproven technol- ogy against smaller but more accessible gains made by applying current best in class technologies and practices. Accessibility A key benefit of AVs is that they may provide transpor- tation for those who are underserved at present and dis- advantaged either by not being able to drive because of impairment or age or not being able to access a car (Shaheen et al. 2017a; Shaheen et al. 2017b; Kroger et al. 2018). These include the elderly, the disabled and low-income communities. If low-cost (or subsidized) transportation services can be enabled by CAVSM, it may improve the livelihood of these populations in several dimensions. Firstly, it could provide better access to employment. In low-density cities with sparse public transport services, a private car can be a critical factor in finding (and keep- ing) paid employment. A recent study by a USC stu- dent (Junken 2015) visualized public transit data from 43 U.S. metropolitan regions (Levinson 2013; Owen and Levinson 2014) to compare the accessibility of work by car compared to transit. For Los Angeles, it was found that 92% of jobs required a public transit commute of greater than one hour whereas only 7% of jobs required a car commute of greater than one hour, with other U.S. urban regions having a similar ratio. Similarly, an analy- sis of the Bay Area in San Francisco illustrated that up to seven times more manufacturing jobs and four times more service jobs were available to residents within a 45-minute car journey than a 45-minute trip using public transit (Golub and Martens 2014). This highlights that car owners have significantly greater access to employ- ment than those who use transit in these metropolitan areas, potentially reinforcing socioeconomic differences between these groups. Grengs (2010) illustrated the importance of automobile ownership in gaining access to jobs for citizens in Detroit, going so far as to recom- mend improving access to automobiles as an effective means of improving employment outcomes for inner city residents. In a similar way, the ability for individuals to access education services might be significantly increased if cost effective CAVSM services can provide appropriate mobility. This could enable residents to stay in school or to attend higher/further education courses such that their employment prospects are significantly increased. In terms of healthcare, improved mobility can enable citizens to attend medical appointments more readily. This would mean that they are less likely to miss school/ work due to more severe illnesses, and it may be possible to treat chronic conditions with outpatient care rather than inpatient care, greatly reducing the overall cost of treatment. This is particularly important in the context of aging populations in the United States and Europe, for whom access to independent mobility becomes chal- lenging in later life (Giesel et al. 2013). Finally, better mobility enabled by CAVSM could enable discretion- ary travel to support attendance at social activities such as family functions or hobbies (Parkhurst et al. 2014) and travel itself as a social activity (Musselwhite 2017), thereby supporting greater societal well-being.

A P P E N D I X A : W H I T E P A P E R 51 Improved mobility could mitigate the negative conse- quences of food deserts. The term ‘food desert’ was first used by a resident of a public sector housing scheme in Scotland in the 1990s to capture the experience of liv- ing in a deprived neighborhood where food was expen- sive and relatively unobtainable (Cummins 2014). Since then the term has grown in popularity to describe urban areas where residents do not have access to healthy and affordable diets. Living in such an area may contrib- ute to social and spatial disparities in health outcomes (Beaulac et al. 2009). In a study of Cologne, Germany (Schneider and Gruber 2013), it was found that residen- tial areas with low income and high deprivation levels had higher availability of unhealthy products (e.g., alcohol, tobacco, fast food) than in affluent neighborhoods. By improving mobility in low-income areas, CAVSM services would enable residents to access a larger number of alter- natives, thereby addressing the diet and diet-related risks. In the United States and the European Union, CAVSM services may bring a significant benefit for rural commu- nities, where the economics for public transport opera- tions are challenging for human-driven vehicles but may become viable for SAVs. Such services would be simi- larly beneficial for non-car users in cities that are more car-oriented (e.g., Los Angeles, Rome), where walking and cycling are difficult and public transport services are sparse. Conversely, in the centers of congested, dense cities with existing high-quality public transport services (e.g., London, New York, Copenhagen), such a service may be counterproductive. While the use of CAVSM to increase access to services can be seen as positive for the affected communities, there is an associated risk. It is established that when the accessibility of a region is improved, the desirability of that region increases with an associated increase in prop- erty values (e.g., Rosiers et al. 2010; Diaz and McLean 1999; Forrest et al. 1996). Those who live within these regions and are unable to stay due to increasing prices may have to move to a lower cost area where the access challenges are worse. This may further marginalize low- income residents, pushing the issues of mobility for low- income neighborhoods further away from city centers and potentially exacerbating the problem. The potential influence of CAVSM on land use and property prices should be overseen in this context. Inclusion An important equity issue is the extent to which trans- port services enable those with additional travel needs, such as the disabled and/or elderly, to satisfy their mobil- ity requirements. With the elderly being more dependent on car use, it is important to note that cessation of driv- ing is predicted by older age, female gender, vision and hearing problems, poor cognitive and physical function- ing, low socioeconomic status, and nursing home place- ment (Anstey et al. 2006; Edwards et al. 2008; Freeman et al. 2006; Gallo et al. 1999). As discussed earlier, being able to travel can make the difference between being able to gain further education, to find/maintain employment, or to attend medical appointments. Connectivity may help in gathering data to understand who has additional travel needs and precisely what those needs are, as well as to know from where they wish to travel and to where they wish to go. On-demand CAVSM services may then be tailored to address these mobility requirements in ways that are not practical today, accommodating differences in physical, mental and sensory ability and technology awareness. An example project, Insight, addressing this issue is developing automated vehicles that are accessible by visually impaired users (http://insight-cav.com/). On the other hand, data emanating from connectiv- ity may lead to inequitable outcomes. Analysis of these data could enable CAVSM service operators to achieve greater depth in their understanding of which vehicles and which passengers tend to use which routes over the course of hours, days, weeks, months and years. The results of such analyses could enable AV operators to determine effectively which routes their vehicles should use. Operators may choose to price routes differentially, depending on willingness to pay. Opting to take a cheaper route may take passenger or freight road users on longer journeys whereas those willing to pay more could access faster ‘premium’ routes with better roads and more reliable journey times. A corollary impact of this differ- ential pricing may be an increase in traffic volumes being routed through low-income neighborhoods, with associ- ated congestion and collision risk. Consequently, it will be important for city authorities to understand how such services and pricing models impact mobility across their networks and to have the regulatory power to influence how such services are operated in order to manage equity and inclusion appropriately. In addition, one of the biggest barriers to use of on- demand CAVSM among populations with additional travel needs is the lack of a smartphone and/or bank card (National Association of Counties 2017). Many new passenger services depend fundamentally on the use of internet connectivity, often through a smartphone, which can exchange important data between the user and the operator. Such data may include personal data about the user, their location and confirmation that they can pay for the mobility service via an electronic payment service. This significantly restricts access to such mobility ser- vices to those who have access to the necessary hardware (smartphone) and the right data package, the ability to download/operate mobile applications that can use elec- tronic forms of payment and those that are comfortable using a smartphone. Smartphone users in both Western Europe and the United States represent just less than 70% of the population (Statista 2017 2018). Clearly

5 2 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S then, mobility services which depend fundamentally on a smartphone would not be accessible to more than 30% of these populations. In high-income European countries and the United States, around 93% of individuals aged over 15 years have bank accounts; however, in lower-income European countries, this figure falls to around 60% (Demirguc-Kunt et al. 2018). In all cases, it is the younger, less-educated and out of work populations that are over- represented in the unbanked group. An even larger propor- tion of the population is considered to be underbanked: individuals who may have an account but lack access to or choose not to access mainstream financial services. Con- sequently, a significant number of potential mobility users could be deprived by inadequate banking facilities, with younger, undereducated and unemployed individuals most at risk. This creates challenging societal inequity through poor access to mobility services. Options to address this problem include the use of prepaid cards, distributing vouchers to trial services to encourage uptake, managing payment for mobility services through housing payments, and providing cash payment options (Serebrin 2016). The equity of service provision–who gets served and at what cost–is a significant issue in the proliferation of SAVs. These services have the potential to be run at much greater efficiency than is achievable today (Fagnant and Kockelman 2018; Greenblatt and Shaheen 2015). By increasing vehicle use and operating costs, profitability can increase. If profitability becomes the key priority, SAV services may be less likely to serve rural, less dense, and some low-income neighborhoods. To avoid facing the challenge of low-demand, such services tend to start in places likely to support highest usage–those with a suf- ficient density of people and uses. Further, as discussed previously, it may be that roads authorities apply tariffs to manage the potential increase in VMT/VKT on their networks. This creates an opportunity and a challenge: to manage the equitable provision of mobility across a city. The distribution of tariffs applied by the authority may allow low-income neighborhoods that are poorly served by public transportation to be served by mobil- ity services at lower costs whereas areas with good quality public transportation links and infrastructure that supports active travel may have higher tariffs on new mobility services to limit their use (Shaheen and Cohen 2018). Public Transit Effects One characteristic of some forms of shared mobility ser- vice is that they can blur the lines between private and public transportation systems. For example, in 2017, the urban navigation and mapping company, CityMapper, introduced a ‘SmartRide’ service to London, its website describing the service as “a hybrid between a bus and a cab,” “a real-time, demand-responsive service,” and “complementing the existing transport infrastructure” (see https://citymapper.com/smartride). Users register for a service on a mobile application and can then request a ride by specifying their start and end locations, using the CityMapper journey planning mobile application. The application then informs the user where to go in order to get picked up by a CityMapper vehicle and where they will be dropped off near to their chosen destination. The operated services were developed based on data col- lected on the movements of individuals across the city and resulted in service routes that could satisfy their predicted unmet demand for mobility at these locations. This approach mirrors similar services such as Chariot (backed by Ford, operating in San Francisco, Austin, and London) and Via (operating in Chicago, New York, and Washington D.C.). A challenge for these new services is how they will integrate with existing public transportation services and projects are examining this concept (e.g., RAMONA project in Berlin, Germany: http://www.dlr.de/vf/ desktopdefault.aspx/tabid-2974/1445_read-50061/ and MERGE project in London, UK: https://mergegreenwich. com/). In Europe and in the United States, these public transportation services are tightly regulated, often with a requirement to serve routes/areas that are typically unprofitable. There is a risk that public authorities could gradually cede greater responsibility for transit services to third parties that would focus on profitability over service to the community, potentially resulting in rein- forcement of inequities. However, it would appear that the disruption caused by such private services is limited to high end transportation business rather than for the working class, while the profitability of ridesourcing ser- vices such as Uber and Lyft remains challenging (Walker 2018). The development of innovative mobility services, especially when empowered by CAV technology, is likely to create opportunities to support mobility in ways that have previously proven to be impractical; however, it is likely that these will need to be monitored and regulated to ensure that their introduction helps to support mobil- ity across the full spectrum of need. Air Quality One of the ways in which city residents can experience inequity is in the quality of the air that they breathe. Dense urban districts with heavier traffic flow are at greater risk of higher concentrations of pollutant emis- sions including nitrogen oxides and particulates. Low- income and minority populations tend to live closer to major roads (Gunier et al. 2003) while the effects of traffic-related pollutants are known to be greater for low-income individuals (Meng et al. 2008, Espino et al. 2015). The shift toward smaller, lower emission vehicles for goods distribution and passenger transport may be

A P P E N D I X A : W H I T E P A P E R 53 facilitated by CAVSM technologies since the duty cycle for an urban transportation vehicle can be comfortably fulfilled by an electric vehicle. Cities can potentiate this change by adopting regulations that encourage the use of lower emission vehicles. By example, London’s Ultra Low Emission Zone (to be launched in April 2019), will only permit vehicles that surpass certain emission stan- dards to enter a central city zone. It should be noted however that addressing the air quality issue as a symp- tom of inequity is not necessarily addressing the cause of inequity itself. This section of the paper has identified a number of ways in which CAVSM services might benefit citizens and businesses. However, the ways in which these ben- efits are distributed will be heavily influenced by the regulatory environment into which they are deployed. As such, regulators can choose to offer minimal inter- vention that may result in a wider array of innovative services being developed but that does not necessarily serve the best interests of all segments of society and may present higher safety risks. Alternatively, they may be more prescriptive over the introduction of CAVSM services perhaps resulting in constraints on innovation in transport provision but with a more defined vision over how such services should be used to improve safety and mobility for city residents and businesses. Given the potential benefits at stake and the safety concerns asso- ciated with transport operations, it is the influence of policymakers in public office in the latter approach that would seem to preferable over the former laissez-faire regulatory approach. dIScuSSIon Connectivity in the transport network has led to the emergence of services that have had a major impact on mobility. It seems likely that automation of road trans- port services will cause a similar or perhaps even greater transformation. Given the huge investments being made in the sector, it has frequently been stated that the intro- duction of AVs is not a question of “if” but “when.” However, the potential for socioeconomic change result- ing from their introduction, means that questions of “where,” “how,” and “why” have equal importance. The uncertainty around the answers to these questions means that the associated socioeconomic impacts of connected and automated vehicles is difficult to esti- mate. In this paper, we have attempted to characterize some of the risks and opportunities that are emerging around this potential impact. Data Privacy and Access CAVSM data enable individuals to be located in a specific space and time. These data contribute to opportunities for greater societal benefits (i.e., increased accessibil- ity and mobility, harmonized traffic flows, innovative mobility services). At the same time, the more detailed the spatial location, temporal position, or individual information included in the data, the more privacy sensitive the data are and the greater the privacy risk. Privacy regulation is much more protective in the Euro- pean Union than in the United States, particularly with the GDPR that is now in effect. Issues connected to open data, data sharing, and data ownership that are all highly associated with CAVSM have the potential to increase privacy risk, which is the likelihood of a privacy problem occurring and the potential magnitude of harm arising from the privacy problem. Linking the degree to which data access is controlled (i.e., greatest ease of use of data to greatest privacy protection) is important for mitigating negative societal impacts from misuse or mis- treatment of personal information. Safety and Security Safety often refers to road traffic safety. When people drive a vehicle, they not only increase their own risk of a crash but also increase crash risks for other motorists, as well as pedestrians and bicyclists. This consequence reflects the social cost of driving, which includes spillover effects on the rest of society such as congestion costs, net output losses, and hospital costs. CAVSM has the opportunity to mitigate as well as exacerbate safety risks. Because more than 90% of traffic crashes are attributed to human error, CAV may greatly reduce these types of error. But CAV may introduce new types of errors such as those resulting from premature release of hardware or software, weather-related technology failures, and cyber- security attacks. Testing on public roads (supported by legislation, funding, and government oversight) is criti- cal for the development of safe operation of CAV. Many such demonstration and large-scale tests are happening in the European Union and in the United States. Economics and Workforce Issues Prosperity is enabled by access to opportunities; trans- portation enables greater access to opportunities. New mobility services that use connectivity (current) and automation (future) are increasing transportation options for passengers and freight. Consequently, there is the opportunity for such services to increase pros- perity across European and U.S. societies. However, as such services emerge, it will be important to be mind- ful of other changes that they may bring. In particular, automation of the driving task may remove a significant source of employment for otherwise low-skilled workers. The predicted benefits of safety and efficiency that accrue

5 4 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S from automation mean that this should not be used to delay its introduction. However, consideration as to how this transition is to be managed will be important. Similarly, changes to mobility may bring about changes in working styles. Time freed up from driving may be used productively for some parts of a journey. This may have a significant impact on land use values and urban planning. Again, it will be for regulatory authorities to monitor these developments and ensure that such sec- ondary effects of transportation are managed to benefit society. Equity Issues around the impact of CVs and AVs on society lie at the heart of how these technologies could bring about changes to social equity. It is early in the implementation of these technologies to state conclusively the full extent of the positive benefit on road safety they may have; how- ever, that such benefits may be distributed more quickly and more equitably as we transition to shared, auto- mated vehicles is a distinct possibility. Similarly, there is the potential to increase mobility options for those who at present have fewer, such as the elderly, the disabled, and poorer communities. However, the interests of these communities will need to be represented in the regula- tory regimes that emerge to manage the deployment of connected and automated services to ensure that these benefits are achieved. The emergence of AVs represents one implementation of a wider issue around the deployment of data-rich, artificially intelligent systems. However, the use of this technology for road transport has important fundamen- tal differences from AI applications that are restricted to digital environments. First, an error in AI that affects the performance of an online service or a smartphone app can be frustrating. An error in AI that affects the performance of an automated vehicle could result in the injury or death of vehicle occupants and/or third par- ties. Fatalities involving vehicles capable of some level of automated driving have already generated global cover- age and reams of commentary, even before official inves- tigators have been able to establish objectively the causes of the crashes. Further such fatalities will occur. To build public trust in connected and automated vehicles, the transportation community must be very clear about the safeguards associated with their testing that are in place to protect the public, why the development of automation is being permitted in public environments and the ethical basis of the decision to proceed with AV development. Well considered, evidence-led socioeconomic arguments may be significant in building the most persuasive case. Second, AVs are being deployed into an environment that is already highly regulated, where the ownership and operation of the underpinning infrastructure is clear and where vehicles will encounter an infinite variety of connected/unconnected and automated/non-automated road users. The strategy adopted by some technology companies to ‘land and expand’ can empower users with remarkable new services but can also harm existing resi- dents and businesses in unpredictable ways. National, regional and local regulators and infrastructure opera- tors can all exert influence over how our roads are used in relation to all of the major themes explored within this paper. It is vital therefore that those in positions of authority engage with new technologies, understand how they might impact upon society and deliver a regulatory environment that maximizes the positive outcomes that may be achieved by the use of CVs and AVs. Socioeco- nomic considerations will be crucial in determining an optimal regulatory response to these technologies. Connectivity and automation are providing the plat- form for some radical changes in transportation and mobility in the United States, in Europe and globally. The rush to bring technologies to market, the uncertain- ties around the impacts of potential deployments, and the challenges in effectively operating existing trans- portation systems mean that it can be challenging for transport regulators to determine the correct response. In these circumstances, a proposed approach would be for regulators not to decide prescriptively what forms of transportation should be used on their networks but to work collaboratively with the industrial and research sectors and engage with their communities to set out ambitious goals for mobility that account for socio- economic issues, including data, safety, security, econom- ics, employment, and equity considerations. Developers of CV and AV technologies would then be incentivized to develop services that match these ambitions. In addition to any direct engagement, the wider public would have a democratic influence over how these transportation priorities are determined. The appropriate management of interests across the public and private sectors will be vital in attaining the potential benefits (and risks) of con- nectivity and automation in our transportation system. referenceS Ackerman, E. (2016). “After Mastering Singapore’s Streets, NuTonomy’s Robo-taxis Are Poised to Take On New Cities.” IEEE Spectrum. Online. Retrieved May 2018 from https://spectrum.ieee.org/transportation/self-driving/ after-mastering-singapores-streets-nutonomys-robotaxis- are-poised-to-take-on-new-cities. Alonso Raposo, M., M. Grosso, J. Després, E. Fernández Macías, C. Galassi, A. Krasenbrink, J. Krause, L. Levati, A. Mourtzouchou, B. Saveyn, C. Thiel, and B. Ciuffo. (2018) An analysis of possible socioeconomic effects of a Cooperative, Connected and Automated Mobility

A P P E N D I X A : W H I T E P A P E R 55 (CCAM) in Europe. Retrieved June 2018 from http:// publications.jrc.ec.europa.eu/repository/bitstream/ JRC111477/kjna29226enn.pdf. Alliance of Automobile Manufacturers and Association of Global Automakers (2014). Consumer Privacy Protection Principles: Privacy Principles for Vehicle Technologies and Services. Online. Retrieved April 2018 from https:// www.globalautomakers.org/media/papers-and-reports/ privacy-principles-for-vehicle-technologies-and-services. Anderson, J. M., N. Kalra, K. Stanley, P. Sorensen, C. Samaras, and O. Oluwatola. Autonomous Vehicle Technology— A Guide for Policymakers. RAND Corporation, Santa Monica, Calif., 2014. Anstey, K., T. Windsor, M. Luszcz, and G. Andrews. Predict- ing driving cessation over 5 years in older adults: Psycho- logical well-being and cognitive competence are stronger predictors than physical health. Journal of the American Geriatrics Society, Vol. 54, No. 1, 2006, pp. 121–126. https://doi.org/10.1111/j.1532-5415.2005.00471.x. Arntz, M., T. Gregory, and U. Zierahn. (2016). The Risk of Automation for Jobs in OECD Countries: A Compara- tive Analysis. OECD Social, Employment and Migration Working Papers, No. 189, OECD Publishing, Paris. http:// dx.doi.org/10.1787/5jlz9h56dvq7-en. Bajaj, V. (2018). The Bright, Shiny Distraction of Self-Driving Cars. Retrieved May 2018 from https://www.nytimes. com/2018/03/31/opinion/distraction-self-driving-cars. html. Beede, D., R. Powers, and C. Ingram. (2017) The Employ- ment Impact of Autonomous Vehicles. U.S. Department of Commerce, Economics and Statistics Administration, Office of the Chief Economist. Retrieved May 2018 from http://www.esa.doc.gov/sites/default/files/Employ- ment%20Impact%20Autonomous%20Vehicles_0.pdf. Beaulac, J., E. Kristjansson, and S. Cummins. A systematic review of food deserts, 1966–2007. Preventing Chronic Disease, Vol. 6, No. 3, 2009. Bhuiyan, J. (2017). A federal agency says an overreliance on Tesla’s Autopilot contributed to a fatal crash. Retrieved May 2018 from https://www.recode.net/2017/9/12/16294510/ fatal-tesla-crash-self-driving-elon-musk-autopilot. Bishop, R. Intelligent vehicle applications worldwide. IEEE Intelligent Systems & Their Applications, Vol. 15, No. 1, 2000, pp. 78–81. https://doi.org/10.1109/5254.820333. Blanco, M., J. Atwood, S. Russell, T. Trimble, J. McClafferty, and M. Perez. (2016). Automated Vehicle Crash Rate Comparison Using Naturalistic Data. https://doi.org/ 10.13140/RG.2.1.2336.1048. Bloom, C. (2017). Self-Driving Cars and Data Collection: Privacy Perceptions of Networked Autonomous Vehi- cles. Retrieved May 2018 from https://www.usenix.org/ conference/soups2017/technical-sessions/presentation/ bloom. Bösch, P. M., F. Becker, H. Becker, and K. W. Axhausen. (2017). Cost-based analysis of autonomous mobility ser- vices. Transport Policy. Retrieved May 2018 from https:// www.ethz.ch/content/dam/ethz/special-interest/baug/ivt/ ivt-dam/vpl/reports/1201-1300/ab1225.pdf. Boston Consulting Group. 2015. Revolution Versus Regu- lation: The Make-or-Break Questions about Autono- mous Vehicles. Online at https://www.bcgperspectives. com/content/articles/automotive-revolution-versus- regulation-make-break-questions-autonomous-vehicles/ ?chapter=3#chapter3_section2. Boudette, B., and E. Vlasik. 2017. Tesla Self-Driving Sys- tem Faulted by Safety Agency in Crash. The New York Times. Retrieved May 2018 from https://www.nytimes. com/2017/09/12/business/self-driving-cars.html. Brooks, S., and E. Nadeau. 2015. Privacy Risk Management for Federal Information Systems. Report 18, No. 8062. Retrieved April 2018 from http://csrc.nist.gov/publications/ drafts/nistir-8062/nistir_8062_draft.pdf. Bryans, J. 2017. The Internet of Automotive Things: vulner- abilities, risks and policy implications. https://doi.org/ 10.1080/23738871.2017.1360926. Journal of Cyber Policy, Vol. 2, No. 2, pp. 185–194. DOI: 10.1080/23738871.2017.1360926. Caballini, C., I. Rebecchi, and S. Sacone. 2017. Maximizing road carriers profit by combining trips and sizing the car- rier coalition. Service Operations and Logistics, and Infor- matics, 2017, IEEE International Conference, pp. 102-107. https://doi.org/10.1109/SOLI.2017.8120978. Canis, B., and D. Peterman. “Black Boxes” in Passenger Vehi- cles: Policy Issues. Congressional Research Service, Wash- ington, DC, 2014. Chan, E., P. Gilhead, P. Jelinek, P. Krejci, and T. Robinson. 2012. Cooperative control of SARTRE automated pla- toon vehicles. 19th ITS World Congress Proceedings. Chester, M., A. Fraser, J. Matute, C. Flower, and R. Pendyala. Parking infrastructure: A constraint on or opportunity for urban redevelopment? A study of Los Angeles County parking supply and growth. Journal of the American Planning Association, Vol. 81, No. 4, 2015, pp. 268–286. https://doi.org/10.1080/01944363.2015.1092879. Christie, D., A. Koymans, T. Chanard, J. M. Lasgouttes, and V. Kaufmann. Pioneering driverless electric vehicles in Europe: The City Automated Transport System (CATS). Transportation Research Procedia, Vol. 13, 2016, pp. 30–39. https://doi.org/10.1016/j.trpro.2016.05.004. Clewlow, R., and G. Mishra. 2017. Disruptive Transporta- tion: The Adoption, Utilization, and Impacts of Ride- Hailing in the United States. Retrieved May 2018 from https://2017_UCD-ITS-RR-17-07%20(1).pdf. Costello, B. Truck driver shortage analysis 2017. The Ameri- can Trucking Association, Arlington, Va., 2017. Cummins, S. Food deserts. The Wiley Blackwell Encyclopedia of Health, Illness, Behavior, and Society, 2014. https:// doi.org/10.1002/9781118410868.wbehibs450. Demirguc-Kunt, A., L. Klapper, D. Singer, S. Ansar, and J. Hess. 2018 Global Findex Database 2017: Measuring Financial

5 6 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S Inclusion around the World and the Fintech Revolution. World Bank Policy Research Working Paper. Retrieved May 2018 from https://globalfindex.worldbank.org/. https://doi.org/10.1596/978-1-4648-1259-0. Dennis, E., and A. Spulber. 2017. International Scan of Connected and Automated Vehicle Technology Deploy- ment Efforts. Retrieved April 2018 from https:// www.cargroup.org/wp-content/uploads/2017/04/CAV_ International_Survey_2017_555402_7.pdf. Diamond, C. C., F. Mostashari, and C. Shirky. 2009. Collect- ing and sharing data for population health: a new para- digm. Health Affairs, Vol. 28, No. 2, pp. 454–466. https:// doi.org/10.1377/hlthaff.28.2.454. Diaz, R. B., and V. A. Mclean. 1999. Impacts of rail transit on property values. American Public Transit Association Rapid Transit Conference Proceedings, pp. 1–8. Diels, C., and J. Bos. Self-driving carsickness. Applied Ergononomics, Vol. 53. Part B., 2016. Dorri, A., M. Steger, S. S. Kanhere, and R. Jurdak. Block- chain: A distributed solution to automotive security and privacy. IEEE Communications Magazine, Vol. 55, No. 12, 2017, pp. 119–125. https://doi.org/10.1109/ MCOM.2017.1700879. Eddington, R. 2006. The Eddington Transport Study. Main Report: Transport’s Role in Sustaining the UK’s Produc- tivity and Competitiveness. Edlin, A., and P. Karaca-Mandic. The Accident Externality from Driving. Journal of Political Economy, Vol. 114, No. 5, 2006, pp. 931–955. https://doi.org/10.1086/508030. Edwards, J. D., L. A. Ross, M. L. Ackerman, B. J. Small, K. K. Ball, S. Bradley, and J. E. Dodson. Longitudinal predictors of driving cessation among older adults from the ACTIVE clinical trial. Journals of Gerontology. Series B, Psycho- logical Sciences and Social Sciences, Vol. 63B, No. 1, 2008, pp. P6–P12. https://doi.org/10.1093/geronb/63.1.P6. Elgart, Z., E. Shipp, J. Cardenas, T. Hansen, and A. Pant. Role of Transportation Network Companies (TNCs) in Reducing Alcohol-Impaired Driving: An Overview. Texas A&M Transportation Institute, College Station, 2016. Espino, J., V. Truong, and E. E. Director. 2015. Electric Car- sharing in Underserved Communities. The Greenlining Institute. Retrieved May 2018 from http://greenlining. org/wp-content/uploads/2015/01/Electric-Carsharing-in- Underserved-Communities-spreads.pdf. Etherington, D. 2018. Aptiv and Lyft’s self-driving BMWs are picking up CES passengers. Retrieved April 2018 from https://techcrunch.com/2018/01/08/aptiv-and-lyfts-self- driving-bmws-are-picking-up-ces-passengers/. ETSC. 2016. Prioritising the Safety Potential of Automated Driving in Europe. Retrieved April 2018 from https://etsc.eu/ wp-content/uploads/2016_automated_driving_briefing_ final.pdf. European Commission. 2006. Regulation (EC) No 561/2006 of the European Parliament and of the Council of 15 March 2006 on the harmonisation of certain social legislation relating to road transport and amend- ing Council Regulations (EEC) No 3821/85 and (EC) No 2135/98 and repealing Council Regulation (EEC) No 3820/85. Retrieved April 2018 from http://eur-lex. europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX: 02006R0561-20150302&from=EN European Commission. 2009. Regulation (EC) No 661/2009 of the European Parliament and of the Council of 13 July 2009 concerning type-approval requirements for the gen- eral safety of motor vehicles, their trailers and systems, components and separate technical units intended there- fore. Retrieved April 2018 from http://eur-lex.europa.eu/ legal-content/EN/TXT/PDF/?uri=CELEX:32009R0661& from=EN. European Commission. 2015. Data Protection Report. Spe- cial Eurobarometer 431/Wave EB83.1 – TNS opinion & social. Retrieved April 2018 from http://ec.europa.eu/ commfrontoffice/publicopinion/archives/ebs/ebs_ 431_en.pdf. European Commission. 2017a. 2016 road safety statistics: What is behind the figures? Retrieved April 2018 from http:// europa.eu/rapid/press-release_MEMO-17-675_en.htm. European Commission. 2017b. Public Support Measures for Connected and Automated Driving. Final Report. Writ- ten by Sociedade Portuguesa de Inovacao (SPI), VTT, and ECORYS. Online. Retrieved April 2018 from http:// edz.bib.uni-mannheim.de/daten/edz-h/gdb/17/CAD%20 -%20Final%20Report%202017.05.31.pdf. European Commission. 2018. On the road to automated mobility: An EU strategy for mobility of the future. Com- munication from the Commission to the European Par- liament, the Council, the European Economic and Social Committee, the committee of the Regions. Retrieved May 2018 from https://ec.europa.eu/transport/sites/transport/ files/3rd-mobility-pack/com20180283_en.pdf. Fagnant, D., K. Kockelman, and P. Bansal. Operations of Shared Autonomous Vehicle Fleet for Austin, Texas, Market. Transportation Research Record: Journal of the Transportation Research Board, No. 2536, 2015, pp. 98–106. https://doi.org/10.3141/2536-12. Fagnant, D. J., and K. M. Kockelman. Dynamic ride-sharing and fleet sizing for a system of shared autonomous vehicles in Austin, Texas. Transportation, Vol. 45, No. 1, 2018, pp. 143–158. https://doi.org/10.1007/s11116-016-9729-z. Favarò, F. M., N. Nader, S. O. Eurich, M. Tripp, and N. Varadaraju. Examining accident reports involving auton- omous vehicles in California. PLoS One, Vol. 12, No. 9, 2017, p. e0184952. https://doi.org/10.1371/journal. pone.0184952. Feigon, S., and C. Murphy. 2016. TCRP Research Report 188: Shared Mobility and the Transformation of Public Transit. Transportation Research Board, Washington, D.C. https:// www.nap.edu/read/23578/chapter/1 Fildes, B., M. Keall, N. Bos, A. Lie, Y. Page, C. Pastor, L. Pennisi, M. Rizzi, P. Thomas, and C. Tingvall. 2015. Effec- tiveness of low speed autonomous emergency braking in

A P P E N D I X A : W H I T E P A P E R 57 real-world rear-end crashes. Accident Analysis and Preven- tion, Vol. 81. Retrieved June 2018 from https://www. sciencedirect.com/science/article/pii/S0001457515001116. https://doi.org/10.1016/j.aap.2015.03.029. Fitzpatrick, A. 2016. Watch Uber’s Self-Driving Trucks Make a Beer Run. Retrieved April 2016 from http://time.com/ 4544135/uber-otto-self-driving-trucks-budweiser-beer/. FMCSA. 2011. Hours of Service of Drivers. Docket No. FMCSA–2004–19608. Retrieved April 2018 from https:// www.gpo.gov/fdsys/pkg/FR-2011-12-27/pdf/2011- 32696.pdf. Forde, C., M. Stuart, S. Joyce, L. Oliver, D. Valizade, G. Alberti, K. Hardy, V. Trappmann, C. Umney, and C. Carson. 2017. The Social Protection of Workers in the Platform Economy. IP/A/EMPL/2016-11 Report for European Parliament’s Committee on Employment and Social Affairs. Retrieved May 2018 from http://www. europarl.europa.eu/RegData/etudes/STUD/2017/614184/ IPOL_STU(2017)614184_EN.pdf. Forrest, D., J. Glen, and R. Ward. The impact of a light rail sys- tem on the structure of house prices: A hedonic longitu- dinal study. Journal of Transport Economics and Policy, 1996, pp. 15–29. Franckx, L. 2017. Shared Mobility and Big Data. Retrieved April 2018 from https://mobilitybehaviour.eu/2017/07/26/ shared-mobility-and-big-data/. Freeman, E., S. Gange, B. Munoz, and S. West. Driving sta- tus and risk of entry into long-term care in older adults. American Journal of Public Health, Vol. 96, No. 7, 2006, pp. 1254–1259. https://doi.org/10.2105/AJPH.2005. 069146. Frey, C. B., and M. Osborne. 2013. The future of employment. How susceptible are jobs to computerisation. Oxford Martin Programme on Technology and Employment. Gallo, J. J., G. W. Rebok, and S. E. Lesikar. The Driving Habits of Adults Aged 60 Years and Older. Journal of the Ameri- can Geriatrics Society, Vol. 47, No. 3, 1999, pp. 335–341. https://doi.org/10.1111/j.1532-5415.1999.tb02998.x. Garfinkel, S. L. De-identification of personal information. NISTIR, Vol. 8053, 2015, pp. 1–46. Gehrke, S., A. Felix, and T. Reardon. 2018. Fare Choices: A survey of ride-hailing passengers in Metro Boston, Report 1, February, MAPC. Gerrard, B. 2017. Just Eat deliver 1,000th meal in London by robot. Retrieved April 2018 from https://www.telegraph. co.uk/business/2017/08/27/just-eat-deliver-1000th- meal-london/. Giesel, F., K. Köhler, and E. Nowossadeck. Alt und immobil auf dem Land? Bundesgesundheitsblatt, Gesundheitsforsc- hung, Gesundheitsschutz, Vol. 56, No. 10, Oct. 2013, pp. 1418–1424. https://doi.org/10.1007/s00103-013- 1832-0. Golub, A., and K. Martens. Using principles of justice to assess the modal equity of regional transportation plans. Jour- nal of Transport Geography, Vol. 41, 2014, pp. 10–20. https://doi.org/10.1016/j.jtrangeo.2014.07.014. Goodwin, P. 2004. The economic costs of road traffic conges- tion. Rail Freight Group, ESRC Transport Studies Unit, University College London. Goos, M., and A. Manning. Lousy and lovely jobs: The rising polarization of work in Britain. Review of Economics and Statistics, Vol. 89, No. 1, 2007, pp. 118–133. https://doi.org/ 10.1162/rest.89.1.118. Government Accountability Office. 2017. Vehicle Data Pri- vacy. Retrieved May 2018 from https://www.gao.gov/ assets/690/686284.pdf. Government Accountability Office. 2013. Vehicle-to-Vehicle Technologies Expected to Offer Safety Benefits, but a Variety of Deployment Challenges Exist. Retrieved May 2018 from https://www.gao.gov/products/GAO-14-13. Government of the Netherlands. 2016. The Declaration of Amsterdam: Cooperation in the field of connected and automated driving. Retrieved April 2018 from https:// www.regjeringen.no/contentassets/ba7ab6e2a0e14e39ba a77f5b76f59d14/2016-04-08-declaration-of-amsterdam ---final1400661.pdf. Greater London Authority. 2018. Mayor’s Transport Strategy. Retrieved May 2018 from https://www.london.gov.uk/ sites/default/files/mayors-transport-strategy-2018.pdf. Greenblatt, J., and S. Shaheen. Automated Vehicles, On- Demand Mobility, and Environmental Impacts. Curr Sustainable Renewable Energy Rep, 2015. https://doi.org/ 10.1007/s40518-015-0038-5. Grengs, J. Job accessibility and the modal mismatch in Detroit. Journal of Transport Geography, Vol. 18, No. 1, 2010, pp. 42–54. https://doi.org/10.1016/j.jtrangeo.2009.01.012. Grosse-Ophoff, A., S. Hausler, K. Heineke, and T. Moller. 2017. How shared mobility will change the automotive industry. Retrieved May 2018 from https://www.mckinsey.com/ industries/automotive-and-assembly/our-insights/ how-shared-mobility-will-change-the-automotive-industry. Gudmundsson, H. Multi-level governance framework for sustainable urban mobility. Sustainable Urban Mobility: A change in governance (M. Finger, ed.), European Uni- versity Institute, Robert Schuman Centre for Advanced Studies, 2013. Gunier, R. B., A. Hertz, J. Von Behren, and P. Reynolds. Traffic density in California: Socioeconomic and ethnic differences among potentially exposed children. Jour- nal of Exposure Science & Environmental Epidemiol- ogy, Vol. 13, No. 3, 2003, pp. 240–246. https://doi.org/ 10.1038/sj.jea.7500276. Hampshire, R. C., C. Simek, T. Fabusuyi, X. Di, and X. Chen. 2017. Measuring the impact of an unanticipated suspen- sion of ride-sourcing in Austin, Texas. SSRN Scholarly Paper, Rochester, N.Y.: Social Science Research Network, May 31, 2017. https://papers.ssrn.com/abstract=2977969. https://doi.org/10.2139/ssrn.2977969. Harding, G., R. Yoon, J. Fikentscher, C. Doyle, D. Sade, M. Lukuc, J. Simons, and J. Wang. 2014. Vehicle-to-Vehicle Communications: Readiness of V2V Technology for

5 8 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S Application. Report DOT HS 812 014. Retrieved May 2018 from file:///C:/Users/j-zmud/Downloads/Readiness- of-V2V-Technology-for-Application-812014%20(5).pdf. Hartgen, D. T., M. G. Fields, and A. T. Moore. 2009. Gridlock and growth: the effect of traffic congestion on regional economic performance. Reason Foundation, No. 371, Los Angeles, Calif. Helman, S., G. B. Grayson, and A. M. Parkes. 2010. How can we produce safer new drivers? TRL Insight Report (INS005), Crowthorne: Transport Research Laboratory. Hong, Q., R. Wallace, and G. Krueger. 2014. Connected v. Automated Vehicles as Generators of Useful Data. Michigan Department of Transportation, Center for Automotive Research, Leidos, Lansing. Hooper, A., and D. Murray. 2017. An Analysis of the Oper- ational Costs of Trucking: 2017 Update. Retrieved April 2018 from http://atri-online.org/wp-content/ uploads/2017/10/ATRI-Operational-Costs-of-Trucking-2 017-10-2017.pdf. IEEE. 2018. Connected Vehicles: An Unchecked Threat to Driver Safety. IEEE XPLORE Innovation Spotlight. Retrieved May 2018 from http://ieeexplore-spotlight.ieee. org/article/connected-vehicles/. International Transport Forum. 2017. Managing the Transition to Driverless Road Transport. Retrieved June 2018 from https://www.itf-oecd.org/sites/default/files/docs/managing- transition-driverless-road-freight-transport.pdf. Jansson, J. Accident Externality Charges. Journal of Transport Economics and Policy, Vol. 28, 1994. Janssen, C. P., and J. L. Kenemans. 2015. Multitasking in autonomous vehicles: Ready to go? Automotive User Interfaces and Interactive Vehicular Applications: Adjunct Proceedings, No. 5. Jevons, W. S. The coal question: An inquiry concerning the progress of the nation, and the probable exhaustion of the coal-mines. Macmillan, New York, 1865. Jolly, I. 2017. Data protection in United States: overview. Prac- tical Law. Retrieved May 2018 from https://content.next. westlaw.com/6-502-0467?transitionType=Default&firstP age=true&bhcp=1&contextData=(sc.Default). Jones, C., and N. Livingstone. The ‘online high street’ or the high street online? The implications for the urban retail hierarchy. International Review of Retail, Distribution and Consumer Research, Vol. 28, No. 1, 2018. https:// doi.org/10.1080/09593969.2017.1393441. Junken, R. 2015. Job Accessibility: Cars vs Transit. Retrieved April 2018 from https://transportationist.files.wordpress. com/2015/09/job-accessibility-revised1.png. Kalra, N., and D. G. Groves. The Enemy of Good: Estimat- ing the Cost of Waiting for Nearly Perfect Automated Vehicles. Rand Corporation, Santa Monica, Calif., 2017. https://doi.org/10.7249/RR2150. Karasin, E. 2017. Tesco delivers groceries in under an hour by ROBOT: Supermarket giant plans to send shopping out on driverless six-wheeled buggies after successful trial. Retrieved April 2018 from http://www.dailymail.co.uk/ news/article-4559162/Tesco-delivers-groceries-hour- ROBOT.html. Kelly Blue Book. 2016. Future Autonomous Vehicle Study. Retrieved April 2018 from https://mediaroom.kbb.com/ future-autonomous-vehicle-driver-study. Kitayama, H., S. Munetoh, K. Ohnishi, N. Uramoto, and Y. Watanabe. Advanced security and privacy in connected vehicles. IBM Journal of Research and Development, Vol. 58, No. 1, 2014, pp. 7:1. https://doi.org/10.1147/ JRD.2013.2288061. Koch, J. 2006. Event Data Recorders and Their Role in Auto- mobile Accident Litigation. Retrieved May 2018 from http://www.jlolaw.com/wp-content/uploads/2015/07/ AutoDiagnosticModules.pdf. Kolodge, K. 2017. Hands off? Not Quite. Consumers Fear Technology Failures with Autonomous Vehicles. JD Power U.S. Tech Choice Study. Retrieved April 2018 from http:// www.jdpower.com/press-releases/jd-power-2017-us-tech- choice-study. Korosec, K. 2018. Waymo Early Riders Can Hail Actual Driverless Minivans Now. Retrieved April 2018 from http://fortune. com/2018/03/13/waymo-driverless-minivans-phoenix/. Kroger, L., T. Kuhminoff, and S. Trommer. 2018 Does con- text matter? A comparative study modelling autonomous vehicle impact on travel behaviour for Germany and the USA. Transportation Research Part A. Retrieved May 2018 from https://ac.els-cdn.com/S0965856417301180/ 1-s2.0-S0965856417301180-main.pdf?_tid=b0ab4d96- 625f-42a8-9a70-e7a9bdacf250&acdnat=1528046697_ d155bb32de0e774148fcb7050224d266 https://doi.org/ 10.1016/j.tra.2018.03.033. Kum, H. C., and S. Ahalt. Privacy-by-design: Understanding data access models for secondary data. AMIA Joint Sum- mits on Translational Science Proceedings AMIA Summit on Translational Science, Vol. 2013, 2013, p. 126. Lazarus, J., S. Shaheen, S. E. Young, D. Fagnant, T. Voege, W. Baumgardner, J. Fishelson, and J. Lott. 2018. Shared Automated Mobility and Public Transport. Road Vehicle Automation 4, Springer, Cham., pp. 141–161. Levinson, D. M. Access Across America: Auto 2013 Data. Center for Transportation Studies, University of Minnesota, Minneapolis, 2013. Litman, T. 2016. Parking Management: Strategies, Evalua- tion, and Planning. Retrieved May 2018 from http://www. vtpi.org/park_man.pdf. Madden, M., and L. Rainie. 2015. Americans’ attitudes about privacy, security and surveillance. Pew Research Center. Retrieved April 2018 from http://www.pewinternet.org/ 2015/05/20/americans-attitudes-about-privacy-security- and-surveillance/. Maia, S. C., and A. Meyboom. Understanding the Effects of Autonomous Vehicles on Urban Form. Road Vehicle Automation, Vol. 4, Springer, Cham., 2018, pp. 201–221. https://doi.org/10.1007/978-3-319-60934-8_17.

A P P E N D I X A : W H I T E P A P E R 59 Markey, E. 2015. Tracking and Hacking: Security and Privacy Gaps Put American Drivers at Risk. Report written by the staff of Senator Edward J. Markey (D-Massachusetts). Retrieved May 2018 from https://www.markey.senate.gov/ imo/media/doc/2015-02-06_MarkeyReport-Tracking_ Hacking_CarSecurity%202.pdf. Martens, K. Justice in Transport as Justice in Accessibility: Applying Walzer’s ‘Spheres of Justice’ to the Transport Sec- tor. Transportation, Vol. 39, No. 6, 2012, pp. 1035–1053. https://doi.org/10.1007/s11116-012-9388-7. McCormick, S. 2017. Key Areas of Security Risk for Con- nected Vehicles. Online. Retrieved May 2018 from http:// cj.msu.edu/assets/ICC-2017-McCormick-Key-Areas-of- Security-Risk-for-Connected-Vehicles.pdf. Meng, Y. Y., M. Wilhelm, R. P. Rull, P. English, S. Nathan, and B. Ritz. Are frequent asthma symptoms among low- income individuals related to heavy traffic near homes, vulnerabilities, or both? Annals of Epidemiology, Vol. 18, No. 5, 2008, pp. 343–350. https://doi.org/10.1016/ j.annepidem.2008.01.006. Meszler, D., O. Delgado, F. Rodríguez, and R. Muncrief. 2018. European Heavy-Duty Vehicles: Cost-Effectiveness of Fuel-Efficiency Technologies for Long-Haul Tractor- Trailers in the 2025–2030 Timeframe. Retrieved April 2018 from https://www.theicct.org/sites/default/files/ publications/ICCT_EU-HDV-tech-2025-30_20180116.pdf. Meyer, J., H. Becker, P. M. Bösch, and K. W. Axhausen. Auton- omous vehicles: The next jump in accessibilities? Research in Transportation Economics, Vol. 62, 2017, pp. 80–91. https://doi.org/10.1016/j.retrec.2017.03.005. Miller, C., and C. Valasek. 2015. Remote Exploitation of an Unaltered Passenger Vehicle. Retrieved May 2018 from http://illmatics.com/Remote%20Car%20Hacking.pdf. Moran, M., B. Ettelman, G. Stoeltje, T. Hansen, and A. Pant. 2017. The Policy Implications of Transportation Network Companies. Policy Research Center Report PRC17-17F. College Station: Texas A&M Transportation Institute. Muller, P. O. Transportation and urban form stages in the spa- tial evolution of the American Metropolis. The Geogra- phy of Urban Transportation, 2nd ed. (S. Hanson, ed.), Guilford, New York, 1995. Murphy, R. Introduction to AI robotics. MIT Press, 2000. Musselwhite, C. Exploring the importance of discretion- ary mobility in later life. Working with Older People, Vol. 21, No. 1, 2017, pp. 49–58. https://doi.org/10.1108/ WWOP-12-2016-0038. National Association of Counties. 2017. Preparing Counties for the Future of Transportation: A Spotlight on Trans- portation Network Companies. Retrieved June 2018 from www.naco.org/sites/default/files/documents/Shared%20 Economies_1pgr_07.06.17_v6.pdf. NACTO. 2017. Blueprint for Autonomous Urbanism. Online. Retrieved April 2018 from https://nacto.org/publication/ bau/blueprint-for-autonomous-urbanism/. Najm, W., J. Koopman, J. Smith, and J. Brewer. 2010. Fre- quency of Target Crashes for IntelliDrive Safety Systems. Report. U.S. Department of Transportation, Research and Innovative Technology Administration, John A. Volpe National Transportation Systems Center. Retrieved May 2018 from https://www.fhwa.dot.gov/publications/ research/connectedvehicles/11040/002.cfm. NHTSA. The Economic and Societal Impact of Motor Vehicle Crashes, 2010 (Revised). U.S. DOT, Washington, D.C., 2015a. NHTSA. Critical Reasons for Crashes Investigated in the National Motor Vehicle Crash Causation Survey. U.S. DOT, Washington, D.C., 2015b. NHTSA. 2016. Traffic fatalities up sharply in 2015. Retrieved April 2018 from https://www.nhtsa.gov/press-releases/ traffic-fatalities-sharply-2015. NHTSA. 2017. Automated Driving Systems 2.0: A Vision for Safety. Retrieved May 2018 from https://www. nhtsa.gov/sites/nhtsa.dot.gov/files/documents/13069a- ads2.0_090617_v9a_tag.pdf. National Transport Commission. 2006. National Transport Commission (Road Transport Legislation—Driving Hours Regulations) Regulations 2006. Retrieved April 2018 from https://www.legislation.gov.au/Details/ F2016C00721. Owen, A., and D. M. Levinson. 2014. Access Across America: Transit 2014 Data. Retrieved May 2018 from https:// conservancy.umn.edu/handle/11299/168064. Parmenter, T. 2017. Driverless vehicle technology will go mainstream in 2018, experts claim. Sky News. Retrieved May 2018 from https://news.sky.com/story/driverless- vehicle-technology-will-go-mainstream-in-2018-experts- claim-11190426. Parry, I., M. Walls, and W. Harrington. Automobile Externali- ties and Policies. Journal of Economic Literature, Vol. 45, No. 2, 2007, pp. 373–399. https://doi.org/10.1257/ jel.45.2.373. Pease, K., and B. Preston. Road safety education for young children. British Journal of Educational Psychology, Vol. 37, No. 3, 1967, pp. 305–313. https://doi.org/ 10.1111/j.2044-8279.1967.tb01946.x. PWC & MI Manufacturing Institute. 2018. Industrial Mobil- ity: How autonomous vehicles can change manufacturing. Retrieved April 2018 from https://www.pwc.com/us/en/ industrial-products/publications/assets/pwc-industrial- mobility-and-manufacturing.pdf. Rayle, L., D. Dai, N. Chan, R. Cervero, and S. Shaheen. Just a better taxi? A survey-based comparison of taxis, tran- sit, and ridesourcing services in San Francisco. Transport Policy, Vol. 45, No. C, 2016. https://doi.org/10.1016/ j.tranpol.2015.10.004. Robarts, S. 2015. EasyMile’s driverless bus rolls-out in Singapore and California. New Atlas. Retrieved May 2018 from https://newatlas.com/easymile-ez10-driverless-bus/ 39891/. Rogers, G. 2017. Ahead of the Curb: The Case for Shared Use Mobility (SUM) Zones. Eno Transportation Weekly.

6 0 S O C I O E C O N O M I C I M P A C T S O F A U T O M A T E D A N D C O N N E C T E D V E H I C L E S Retrieved April 2018 from https://www.enotrans.org/ article/ahead-curb-case-shared-use-mobility-sum-zones/. Rosiers, F. D., M. Thériault, M. Voisin, and J. Dubé. Does an improved urban bus service affect house values? Inter- national Journal of Sustainable Transportation, Vol. 4, No. 6, 2010, pp. 321–346. https://doi.org/10.1080/ 15568310903093362. Russo, F., and A. Comi. City characteristics and urban goods movements: A way to environmental Transportation sys- tem in a sustainable city. Procedia: Social and Behavioral Sciences, Vol. 39, 2012, pp. 61–73. https://doi.org/10.1016/ j.sbspro.2012.03.091. Schaller, B. 2017. Unsustainable? The Growth of App-Based Ride Services and Traffic, Travel, and the Future of New York City. Retrieved May 2018 from http://www. schallerconsult.com/rideservices/unsustainable.pdf. Schneider, S., and J. Gruber. Neighbourhood depriva- tion and outlet density for tobacco, alcohol and fast food: First hints of obesogenic and addictive environ- ments in Germany. Public Health Nutrition, Vol. 16, No. 7, 2013, pp. 1168–1177. https://doi.org/10.1017/ S1368980012003321. SCR. 2018. Paratransit Driver Job Description. Retrieved April 2018 from http://scrdrivers.com/job-description/. Sener, I., J. Zmud, and C. Simek. Examining Future Automated Vehicle Usage: A Focus on the Role of Ridehailing. Texas A&M Transportation Institute, College Station, 2016. Serebrin, H. 2016. Improving Unbanked Access to Shared Mobility Services. Capstone Project for Seattle Department of Transportation. Retrieved May 2018 from https://www. slideshare.net/HesterSerebrin/serebrincapstonefinal. Shaheen, S., N. Chan, A. Bansal, and A. Cohen. 2015. Defini- tions, Industry Developments, and Early Understanding. White Paper. Shared Mobility: A Sustainability and Tech- nologies Workshop, Berkeley Transportation Sustainability Research Center, Caltrans. Retrieved May 2018 from http:// innovativemobility.org/wp-content/uploads/2015/11/ SharedMobility_WhitePaper_FINAL.pdf. Shaheen, S., A. Cohen, and I. Zohdy. 2016. Shared Mobil- ity: Current Practices and Guiding Principles. FHWA- HOP-16-022. Retrieved April 2018 from https://ops.fhwa. dot.gov/publications/fhwahop16022/fhwahop16022.pdf. Shaheen, S., C. Bell, A. Cohen, and B. Yelchuru. 2017a. Travel Behavior: Shared Mobility and Transportation Equity. No. PL-18-007. https://www.fhwa.dot.gov/policy/otps/ shared_use_mobility_equity_final.pdf Shaheen, S., A. Cohen, B. Yelchuru, S. Sarkhili, and B. A. Hamilton. 2017b. Mobility on Demand Operational Concept Report. No. FHWA-JPO-18-611. United States Dept. of Trans- portation, ITS Joint Program Office. Shaheen, S., and A. Cohen. Is It Time for a Public Transit Renaissance? Navigating Travel Behavior, Technology, and Business Model Shifts in a Brave New World. Journal of Public Transportation, Vol. 21, No. 1, 2018, pp. 67–81. https://doi.org/10.5038/2375-0901.21.1.8. Shergold, I., M. Wilson, and G. Parkhurst. 2016. The mobility of older people, and the future role of Connected Autono- mous Vehicles. Project Report. Centre for Transport and Society, University of the West of England, Bristol, Bristol. Available from http://eprints.uwe.ac.uk/31998. Shirgaokar, M. 2017. Which Barriers Prevent Seniors from Accessing Transportation Network Company (TNC) Ser- vices? Identifying Ways Forward for a Gendered Policy Approach. Presented at 96th Annual Meeting of the Transportation Research Board, Washington, D.C., 2017. Simonite, T. 2016. Mining 24 Hours a Day with Robots. MIT Review. Retrieved May 2018 from https://www. technologyreview.com/s/603170/mining-24-hours- a-day-with-robots/. Simpson, B. 2018. Peloton predicts commercial launch of truck platooning service this year. Transport Topics. Retrieved May 2018 from http://www.ttnews.com/articles/peloton- promises-commercial-platooning-2018. Smith, A., and M. Anderson. 2017. Americans’ attitudes toward self-driving vehicles. Pew Research Center. Retrieved May 2018 from http://www.pewinternet.org/2017/10/04/ americans-attitudes-toward-driverless-vehicles/ Sotto, L., and A. Simpson. 2014. Data Protection & Privacy 2015, United States, Getting the Deal Through. Retrieved April 2018 from https://www.huntonprivacyblog.com/ wp-content/uploads/sites/18/2011/04/DDP2015_United_ States.pdf. Statista. 2017. Smartphones in the U.S.—Statistics & Facts. Retrieved May 2018 from https://www.statista.com/ topics/2711/us-smartphone-market/. Statista. 2018. Smartphone user penetration as percentage of total population in Western Europe from 2011 to 2018. Retrieved May 2018 from https://www.statista.com/ statistics/203722/smartphone-penetration-per-capita-in- western-europe-since-2000/ Stocker, A., and S. Shaheen. 2016. Shared Automated Vehi- cles: Review of Business Models. Discussion Paper for Roundtable on Cooperative Mobility Systems and Auto- mated Driving, Ottawa, Canada, December. http://www. itf-oecd.org/co-operative-mobility-systems-automated- driving-roundtable. Stocker, A., and S. Shaheen. Shared Automated Mobility: Early Exploration and Potential Impacts. Road Vehicle Auto- mation 4. Springer, Cham, Switzerland, 2017, pp. 125–139. Stocker, A., and S. Shaheen. Forthcoming. Shared Automated Vehicle (SAV) Pilots and Automated Vehicle Policy in the U.S.: Current and Future Developments, Springer. https:// doi.org/10.1007/978-3-319-94896-6_12. The Economist. 2018. Reinventing Wheels. March 1, 2018. Retrieved May 2018 from https://www.economist.com/ special-report/2018/03/01/autonomous-vehicles-are-just- around-the-corner. Todd, S., and W. Waters. 2018. Europe’s leading reefer haulier highlights driver shortage. Retrieved April 2018 from https://www.lloydsloadinglist.com/freight-directory/news/

A P P E N D I X A : W H I T E P A P E R 61 Europe%E2%80%99s-leading-reefer-haulier-highlights- driver-shortage/71128.htm#.Ws_miCjwaUk. Toffler, A., and T. Alvin. The Third Wave, Vol. 484. Bantam Books, New York, 1980. Traffic Impact Newswire. 2016. Road Safety: 26,000 killed in Europe traffic accidents in 2015, social cost €€100 bil- lion. Retrieved April 2018 from https://www.travel-impact- newswire.com/2016/04/road-safety-26000-killed-in- europe-traffic-accidents-in-2015-social-cost-e100-billion/. TRB. Special Report 319: Between Public and Private Mobility: Examining the Rise of Technology Enabled Transportation Services, Transportation Research Board, Washington, D.C., 2016. Trimble, T. E., R. Bishop, J. Morgan, and M. Blanco. 2014. Human Factors Evaluation of Level 2 and Level 3 Auto- mated Driving Concepts: Past Research, State of Automa- tion Technology, and Emerging System Concepts. Report No. DOT-HS-812-043. National Highway Traffic Safety Administration, Washington, D.C. Visser, J., T. Nemoto, and M. Browne. Home delivery and the impacts on urban freight transport: A review. Procedia: Social and Behavioral Sciences, Vol. 125, 2014, pp. 15–27. https://doi.org/10.1016/j.sbspro.2014.01.1452. Wegman, F. 2017. The future of road safety: A worldwide perspective. IATSS Research, Vol. 40, No. 2. Retrieved June 2018 from https://www.sciencedirect.com/science/ article/pii/S0386111216300103. https://doi.org/10.1016/ j.iatssr.2016.05.003. Westin, A. F. Privacy and freedom. Atheneum, New York, 1967, p. 7. Wong, J. 2017. Delivery robots: a revolutionary step or sidewalk-clogging nightmare? The Guardian. Retrieved May 2018 from https://www.theguardian.com/ technology/2017/apr/12/delivery-robots-doordash-yelp- sidewalk-problems. Zhang, W., S. Guhathakurta, J. Fang, and G. Zhang. Explor- ing the impact of shared autonomous vehicles on urban parking demand: An agent-based simulation approach. Sustainable Cities and Society, Vol. 19, 2015, pp. 34–45. https://doi.org/10.1016/j.scs.2015.07.006. Zipper, D. 2018. Who Owns Mobility Data? Retrieved April 2018 from https://www.citylab.com/transportation/ 2018/01/who-owns-urban-mobility-data/549845/. Zmud, J., J. Wagner, M. Moran, and J. P. George. 2016a. License Plate Reader Technology: Transportation Uses and Privacy Risks. NCHRP 08-36: Task 136. Retrieved April 2018 from https://scholarship.law.tamu.edu/cgi/ viewcontent.cgi?article=1920&context=facscholar. Zmud, J., I. Sener, and J. Wagner. 2016b. Consumer Accep- tance and Travel Behavior Impacts of Automated Vehicles. Final Report. PRC-15-49F. Texas A&M Transportation Institute, College Station.