The Fourth Branch project of the Mercatus Center at George Mason University is dedicated to advancing knowledge about the effects of regulation on society, commerce, and innovation. As part of its mission, the program conducts independent legal and economic analyses to assess agency rulemakings and proposals from the perspective of consumers and the public. In the past, it has published research related to the regulatory approach to transportation innovation, including highly automated vehicles (HAVs).
We welcome the opportunity to submit comments regarding the request from General Motors for exemptions from the National Highway Traffic Safety Administration (NHTSA) for automated driving systems. HAVs have the potential to significantly reduce the number of car accidents and provide many other safety benefits. The current flexible approach has focused on enabling technological innovation in this field. With these potential benefits in mind, our comments focus on the following points:
- Regulatory exemptions for HAVs should proceed until existing law is updated.
- NHTSA should be permissive in exempting low-emission vehicles.
- Economic experience suggests that safety benefits from a permissive approach toward HAV deployment.
I. NHTSA’s Current “Soft Law” Approach to HAVs is Prudent, but Regulatory Exemptions Should Proceed Until Existing Requirements are Updated for HAVs
NHTSA has taken a largely “soft law” approach to regulating autonomous vehicles through various guidance documents, working groups, and other informal approaches. Through these various subregulatory approaches, NHTSA’s stated aim is to enable and encourage the development of these technologies along diverse and competing pathways as a way to maximize their social benefit. This regulatory approach thus far has allowed innovators to rapidly develop solutions that they see as being demanded by the market.
While there are many benefits to soft law, as technology advances faster than formal policy solutions, the Department of Transportation (DOT) will have to consider granting exemptions on a case-by-case basis until it reaches a point where regulation is both needed and appropriate, given the state of the technology. Soft law serves as a second-best solution, but it will allow time for legislative bodies and agencies to gain a greater understanding and avoid enacting laws that will limit or dictate the ways in which technology should develop. For emerging technologies, soft law should be preferred over stifling regulation.
Exemptions from certain standards will need to be considered as new technology does not fit into current regulatory boxes but may provide safer or better alternatives. Recently, as part of AV 3.0, the DOT stated that it will no longer presume that the operator of a vehicle is human; however, existing regulations may not quickly adjust to such a change in presumption without the need for exemptions. In fact, an HAV operating system is likely to be safer than a human operator, as discussed later in this comment.
Providing exemptions even in particular cases may provide more general guidance over what kinds of technologies can be anticipated to receive exemptions going forward. Ideally, these exemptions will gradually provide a general framework for safety and design while still encouraging competition in the market. While case-by-case exemptions provide regulatory flexibility, failure to establish more general rules on updated safety or design of HAVs could lead to regulatory uncertainty and have the undesirable effect of slowing down product development. As such, the current exemption method should be considered a stopgap until standards can be updated to reflect innovative advancements.
II. NHTSA Should be Permissive in Exempting Low-Emission Vehicles
NHTSA asks, “What studies, data, assumptions, scientific reasoning, and methodologies are needed for the agency to evaluate and compare the [zero-emission autonomous vehicle] and a [Federal Motor Vehicle Safety Standards]-compliant non-ADS vehicle? For example, should the agency assess whether an ADS steers, brakes, and accelerates at least as effectively and safely (e.g., as quickly) as the average human driver? If so, what methodology should it use? Are there other approaches to making the safety evaluation and comparison? Please provide specific references to all sources of such tools or evaluation approaches.”
When considering exemptions for low-emission motor vehicles, requiring the applicant to show safety at least equal to the safety level of the standard would be too stringent and unsupported by law. Section 30113(b)(3)(B)(iii) includes a somewhat permissive legal standard for evaluating exemption requests for low-emission motor vehicles, and it is certainly less stringent than the standard for exemptions under (B)(ii) and (B)(iv).
Under the statute, NHTSA may provide exemptions to low-emission vehicles with features that would not “unreasonably lower the safety level” of those vehicles. Whether something is unreasonable is obviously subject to differences of opinion. However, it’s notable that forbidding what is unreasonable, as 30113(b)(3)(B)(iii) does, is a fairly permissive legal standard. As legal scholar Benjamin Zipursky noted, such a rule “permits everything except that which is unreasonable” and initiates an inquiry to determine what is excessive.
The question before NHTSA, then, is whether an exemption would excessively lower the safety level of a vehicle. This is more permissive than the standard in (B)(ii) and (B)(iv), which generally requires applicants to show that their exempted vehicle would provide a safety level at least equal to the safety level of the standard.
III. Modeling the General Public Interest and Safety Benefits of a Permissive Approach Toward Highly Automated Vehicles
As Secretary Chao has noted, HAVs have “the potential to impact safety significantly—by reducing crashes caused by human error, including crashes involving impaired or distracted drivers, and saving lives.” Impaired drivers are believed to have been responsible for 3,000 deaths in 2017.
Unlike human drivers, HAVs never get drunk, drowsy, or distracted. Further, HAV computer systems benefit from the crowdsourced nature of HAV development. A hard-to-see pedestrian crosswalk, for instance, would be “known” by every HAV using the same mapping system, whereas the existence of the crosswalk must be relearned by every new human driver who passes by. HAV mapping and sensor systems will likely improve in time. The same cannot be expected from human drivers.
A permissive regulatory approach toward HAVs should be warranted if the benefits of that approach exceed the costs in terms of loss of life. Mercatus scholars have noted in a previous public interest comment to NHTSA that “every policy decision must be judged by how rapidly it can help us advance that goal [of zero traffic deaths] and move us away from the current baseline [100% human drivers] in terms of both lives lost and other costs to citizens and the economy.”
That public interest comment (attached herein) estimates the diffusion rates of HAVs and the accompanying reduction in traffic deaths. From those estimates, the comment models the costs in lives from regulatory delay, finding that a 10 percent delay in HAV deployment, for instance, corresponds to about 34,000 additional deaths during a 30-year diffusion period. A Rand Corporation study similarly finds that delaying the deployment of HAVs could have severe costs in terms of human life, particularly in the long term. The authors state that, for instance, they “could not find a plausible set of conditions in which waiting for significant safety improvements saved more lives in the long run.” The long-term cumulative cost of waiting could be, according to their models, as high as half a million lives lost.
We applaud the NHTSA’s commitment to improving transportation and transportation safety by encouraging innovation in HAVs. The underlying law and the economic research we’ve provided call for a permissive approach when granting exemptions for low-emission HAVs. We hope NHTSA will continue its case-by-case analysis until the regulatory framework is adequately updated.