Friday, October 07, 2016

Self-Driving Cars, Equal Protection, and Federalism

by Michael Dorf

In my latest Verdict column I ask whether non-self-driving cars should eventually be banned. Although the regulatory guidelines recently released by the National Highway Traffic Safety Administration (NHTSA) address a great many questions, they do not take a position on this one--which my column admits is probably about two decades away.

Nonetheless, because progress on self-driving cars has come much more quickly than many experts expected, it is possible that the issue could be live sooner. In any event, it's worth thinking about now because it could have implications for phasing in self-driving cars. For example, some new roads that are appropriate for self-driving cars might not be safe for human drivers. Think about how closely two trains going in opposite directions can pass one another. Self-driving cars that communicate with each other on "smart highways" could likewise use narrower roads than are safe for cars with human drivers.

My column lays out the costs and benefits of banning human-driven cars from public roads and (spoiler alert!) concludes that the benefits of a ban outweigh the costs. If the legal system eventually reaches the same conclusion, driving as we know it could be relegated to amusement parks.

Here I want to point to a constitutional and a quasi-constitutional implication of the shift to self-driving cars.

The constitutional implication concerns equal protection. Widespread use of self-driving cars should greatly reduce opportunities for police to engage in racial profiling. Not to zero, of course. Much racial profiling victimizes pedestrians. But the reduction should be substantial nonetheless.

To see why, consider Prof. Colb's 2001 law review article Stopping a Moving Target. In addition to making other important points, the article takes note of the fact that it is effectively impossible to drive any non-trivial distance without violating some traffic regulation. And because the Supreme Court held in Whren v. United States that a pretextual traffic stop does not violate the Fourth Amendment, Prof. Colb explains that police therefore have carte blanche to engage in racial profiling. Yes, she acknowledges, an officer whose real reason for stopping a car is the race of the driver has violated equal protection, but in practice it is nearly impossible to prove an illicit racial motive. Her proposed solution is to forbid traffic stops except in cases of actual imminent threat to the public. However, the courts have not adopted Prof. Colb's proposal, and so motorists continue to experience traffic stops for "driving while Black"--far too frequently with fatal consequences.

Self-driving cars should substantially reduce this phenomenon for the simple and obvious reason that self-driving cars would be programmed to comply with all applicable traffic laws. Even when, for whatever reason, a fully self-driving car violates a traffic law (perhaps to avert a greater danger), there would be no reasonable suspicion that the driver of the vehicle had committed a traffic offense because there would be no human driver of the vehicle.

To be sure, self-driving cars would not eliminate the opportunities for racial profiling on the roads. Police could say that they stopped a vehicle because one of the passengers matched the description of a suspect in some crime. Or they could say that they suspected that the vehicle was stolen.

But these sorts of claims would be hard for the police to make effectively. Although police sometimes do make the "matched the description" claim, that would be relatively easy to challenge in cases where the supposed match is based simply on the race of a passenger passing by at speed. Meanwhile, in a world of self-driving cars, it is likely that individual car ownership would plummet and robo-taxi companies could readily program anti-theft measures into the vehicles themselves. Stealing a robo-taxi would be effectively impossible and pointless, rendering a vehicle stop on suspicion of car theft highly dubious.

Again, I don't want to sound like a utopian. Self-driving cars will not end racial profiling, much less racism more broadly. But they will help.

* * *

Now onto a quasi-constitutional effect of self-driving cars: federalism. I call this a quasi-constitutional issue in recognition of the fact that Congress has the power to regulate self-driving cars directly or by delegating authority to a federal agency and that Congress could, if it wished, pre-empt state and local regulation of driving. This is not a difficult constitutional question (except perhaps for Justice Thomas). Cars are both articles of commerce and instrumentalities of commerce.

Nonetheless, in recognition of the important role of states in our federal system, Congress does not exercise anything like its full power over automobiles. Federal law delegates to NHTSA authority to implement a variety of safety laws for motor vehicles. Yet it is fair to say that state authorities, exercising their non-pre-empted reserved powers, are the primary actors in regulating automobiles. They license drivers (at least while human drivers still exist) and they set comprehensive rules of the road, including speed limits. True, from 1974 through 1995, federal law prescribed a maximum speed limit of 55 mph on interstate highways, but federalism concerns led to the law's repeal, even at the cost of a three percent increase in highway deaths.

The NHTSA self-driving car guidelines recognize the shared federal/state responsibility for automobile regulation. But one might wonder whether that makes sense. Given the inherent mobility of cars and especially the need for interoperability among the robo-cars of the future, wouldn't we better off with a single national regulatory regime? If technology brings us all closer together and effectively shrinks the world, shouldn't that be the ground for increasing the scope of federal legislation at the expense of state legislation? That, after all, is the standard explanation for the expansion in the effective scope of the Commerce Clause since 1789: As the national economy became more tightly integrated, the domain of the truly local shrank.

That remains a good argument as a constitutional matter. Congress has the power to regulate automobiles and pre-empt state law. But it doesn't follow that Congress should exercise that power. Responding to what Herbert Wechsler famously called the "political safeguards of federalism," Congress has refrained from displacing state law regarding automobiles in deference to state and regional variations in conditions and preferences. The key point I want to make here is that self-driving cars make it easier to accommodate state-by-state and even local variation in automobile regulation.

Suppose you are driving a conventional car from Maine to Florida on I-95. Your route subjects you to the traffic laws of fifteen states plus the District of Columbia. You will be subject to additional local variations when you get off the highway for gas, a meal, lodging, or a bathroom break. I suppose that in theory an especially law-abiding driver will therefore attempt to familiarize himself or herself with all of the different laws, but the vast majority of drivers will simply assume that the traffic law en route is more or less what it is at home. That will be right for a great many matters but not for everything. As a result, if we really wanted to increase human compliance with traffic law, we would favor a single body of national law.

But self-driving cars make federalism concerns a snap. The GPS on the car tells it where it is, and the state and local traffic laws can be programmed in seamlessly.

This strikes me as a part of a larger phenomenon. In the age of heavy industry, integration of the national economy generally recommended uniform national regulation. If you are making a Model A Ford on a production line, you want to be able to make every vehicle the same, because the fixed cost of production is high relative to the marginal cost. Complying with fifty different state regimes is more costly than complying with one uniform federal law--even if the state laws are on average no stricter than the federal one. Likewise, if you are packaging cigarettes using 1960s technology, you want to be able to make every package the same. Hence, the pre-emptive force of the Cigarette Labeling and Advertising Act of 1966.

However, modern production techniques are much more flexible than in the past. That's one of the main reasons why we have much shorter product cycles than we used to. Manufacturers can use the same equipment to make a variety of products, because the equipment can be reprogrammed pretty easily. Of course, there are still special purpose manufacturing tools, but on the whole the direction is towards flexible production. And that trend is likely to continue over time as 3D printing and similar technologies improve.

There are all sorts of implications of the ongoing shift in technology. For example, notwithstanding the lies of Donald Trump, manufacturing in the U.S. has increased over recent decades. Manufacturing jobs have been lost mostly to automation and productivity gains, not to Mexico and China. But even though Trump's diagnosis and prescription are idiotic, the loss of those manufacturing jobs is a real concern.

It is not, however, my concern in this post. My relatively modest point here is that however one tallies up the net costs and benefits of automation, the example of self-driving cars suggests that it does not point towards increasing nationalization of regulation. It actually facilitates federalism.


Joe said...

I can see a few remaining reasons for stops with self-driving cars, putting aside that machine error can result in violations. For instance, issues related to the car alone including a broken headlight (an issue in a recent case) or something.

The federalism issue reminds me of one of the "scary" hypos in the PPACA cases -- what stops Congress from some requirement related to cars, be it insurance or otherwise. I didn't find it that scary -- (1) health insurance was a clearer case, including the scope of the problem addressed (2) various things there DID seem possible. Some requirement to have resources for an accident, for instance, especially given so much of the roadways are closely tied to interstate highways, their service roads or vehicles going to and from them.

As the discussion notes, there is a way to balance federalism concerns, just as PPACA does regarding exchanges and now Medicaid expansion (which could have been in the law itself if Republicans played nice & offered it as a compromise). Also, merely because in theory something can be nationally regulated, it need not be. Political pressures limit things there.

Greg said...

I'm not so sure that states will outlaw non-self-driving cars, particularly in a 20-year kind of timeframe.

There are major problems with the related issues of liability and insurance with cars than cannot be operated by their owners. There are also practical issues. If a car "can" be operated by the owner, then the owner can reasonably be held liable if something goes wrong. If it can't, then the manufacturer is pretty much completely responsible. As a practical matter, when things get "funny" a self-driving car is going to stop and, if possible, pull over. This doesn't square well with rules that require drivers to exit the freeway if their vehicle is still minimally functional. Self-driving cars may require towing more often than human-driven cars in malfunction situations.

However, insurance and federal standards may also provide a way for producing nearly all self-driving cars without actually requiring them.

First, the federal government requires all cars to be built with self-driving capabilities, including some interoperability standards.

Second, the federal government allows automobile insurance carriers to provide insurance that is only valid in self-driving mode or during emergencies. This presumably lower-price insurance would not be available on older cars that did not have a self-driving mode.

Third, the federal government mandates that all cars default to self-driving mode, and provide the following functions:
1.) Provide an "emergency mode" where certain restrictions are enforced, such as a maximum speed of 30mph and emergency flashers activated.
2.) Provide a "user-driven mode" where the user is given control, but that can only be activated by the user proving that they actually have insurance for user-driven mode. Some standardized form of providing this proof would be defined.

Fourth, at some time in the more distant future, the federal government would make the user-driven mode optional, and might even make the emergency mode optional.

Fifth, some roads could be built that are not legal for cars in user-driven mode.

If self-driving cars are truly that much safer, then the cost of user-driven insurance will eventually become so prohibitively expensive (at least for a primary vehicle) that almost no one will carry it. That effectively makes all cars self-driven without actually requiring it.

BTW, I think 20 years is WAY too short of a timeline for this. 10-20 years is probably closer to the "all cars must have self-driving capabilities" part of the sequence. (Which is still much faster than I thought it would be 10 years ago. The insurance and liability issues alone looked completely insurmountable then.)