What to expect as the self-driving vehicle approaches

April 2017 | Bern Grush | Grush Niles Strategic

It is commonly assumed that the vehicles we insure are about to become automated vehicles (AVs). The full commercial and societal implications of this technology are subject to diverse forecasts and interpretations, but there is no doubt that momentous change is coming.

One of the most common predictions is that the great majority of crashes will disappear — although it is uncertain how quickly. Tests have shown that human attention while operating partially-automated vehicles can be difficult to re-acquire after a period of self-driving. While monitoring automated operation, the normally responsible vehicle operator may pay only cursory attention while engaging in other tasks or distractions. This implies a potential for escalating risk during this present early period of vehicle automation. Full and unconstrained automation that needs no human driver backup while achieving these safety promises is much farther away — perhaps decades.

Another commonly held prediction is that few people, if any, will bother to own a personal or family vehicle because swarms of ubiquitous robo-taxis would respond in minutes to a smartphone app. Such a dramatic shift seems tenuous and is certainly not imminent.

While these are only two of many hundreds of AV predictions, they are among the ones that elicit the greatest popular enthusiasm. These are also the source of the greatest concerns among those whose businesses or careers rely on the personal auto insurance industry. The first prediction indicates lower premiums in the long run, while the second suggests fewer policies written. A double hit.

Our research explores whether and how these and many other AV predictions might be realized. How soon would such changes occur? In what order? With what side effects? Can they be accelerated? Will they be delayed?

Stages of AV change

There are three major stages to significant market change based on technological innovation: development, deployment and diffusion. While development relies on technology innovation, deployment and diffusion rely on societal and human behavior. While development and deployment are prerequisites, what matters most to the insurance industry is the diffusion of AVs: When will who be buying how many of what and how will they be insured?

Development

Remarkable development progress in AV technology has been made. The automobile as we now know it is sure to become a casualty under the boot of digital progress — but we are far from ready by the measure of an AV that can go anywhere, anytime without human accompaniment and oversight. If that utopian goal is still a few decades away what then of the interim? What vehicles will be purchased and how will they be insured?

Furthermore, there may never be a time of absolute vehicle autonomy. Even in the more distant highly-automated future, roads and vehicle fleets would be still managed by human overlords, analogous to air-traffic controllers managing our skies today. This is already the case with the first automated mini-buses deployed in several countries. Mobility will not become risk-free where human inattention becomes absolute.

Deployment

Deployment is the second stage, and here it gets more difficult. Regardless of how intelligent vehicle technology becomes, there must be places, management systems and rules for their use. Over decades of growing deployment, these new vehicles are expected to share the road with non-automated cars and taxis, bikes, pedestrians, road-construction, detritus, potholes and snowdrifts. To this mix will be added transit vehicles, delivery trucks and service vehicles — themselves in various states of automation and jockeying for lane and parking space. One easily appreciates why significant loss-history is critical to actuarial calculations before ratemaking responses are significant and applied.

MIT roboticist, David Mindell, asks: “If people are freed from the tedium of driving and are sleeping, or reading a book, how will they rush into the (control) loop quickly enough to avoid oncoming traffic, or a pothole, or a collision?”  While roboticists will say this sort of problem will go away in the long term as the technology is perfected and the driver becomes fully redundant, it is a critical problem for early deployment — so much so that Dr. Nidhi Kalra (Rand Corp) told a U.S. Congressional hearing on AVs: “There is evidence to show that Level 3 (semi-automation) may show an increase in traffic crashes.” 

Consider that however much has been accomplished in the laboratory, in pilots and in trials, so far we have not yet deployed full-automation in the road environments from which property & casualty actuaries harvest their data. Mindell observes: “when lives and resources are at stake, we have reined in the autonomy. It is not a story about progress — that one day we’ll get it right — but a story about the move from a laboratory to field. That transition tempers autonomy…”  Mindell and Kalra’s warnings portend a slower, more difficult and riskier deployment as the promised self-driving cars arrive in manufacturers’ show rooms in the early 2020s.

Diffusion

What is germane to the insurance industry for the next decade or more is that AV technology will diffuse via two sorts of conditional automation — one that a driver can flick on and off at will, and another that is fully reliable without a human driver, but operates only within limited circumstances: locations, lanes, times, and weather conditions.

The first, which we call Market-1, is a steadily improving consumer product. Already available in a few premium AV models, these will be owned, insured and used in ways that match today’s use of personal or household vehicles. These will get more convenient year over year. They will require a driver who will use the controls less and less frequently. They will share the same roads we have today with every other sort of non-automated and conditionally automated vehicle, bus, taxi, bicycle and pedestrian. These personally owned vehicles, as they gradually decline in price, grow in reliability and support a used-vehicle market, will encourage ownership and kilometres-travelled — a growth market.

If this were the entire story, it would be predictable that over a couple decades, the Market-1 vehicle would become the only vehicle available, would gradually replace the current fleet, that eventually the driver would become unnecessary and finally banned. And by that point the projected safety promises and concomitant drop in premiums could be realized.

But this is not the whole story. The second form of vehicle automation, Market-2, is a fully automated but constrained-use vehicle immediately suited for use as robo-taxis, local micro-transit, and first/last kilometre services feeding into existing transit systems. This, too, is just becoming available in limited, slow, easy-to-deploy circumstances. Market-2 vehicles will also become more robust over time, but without a human driver will remain spatially and speed constrained for their first couple of decades.

Given these two major formats for AV deployment, we can predict the nature of AV diffusion. It is the consumer of Market-1 vehicles that will require personal auto insurance for the foreseeable future. (Some Market-1 manufacturers are already disrupting this model, with Tesla bundling lifetime insurance and maintenance costs into the price of their vehicles. ) Market-2 ride-seller vehicles will be insured by commercial fleet insurance, and may even be self-insured by the fleet manufacturers.

Policy volumes, premiums

As AV diffusion dawns, today’s non-automated car owners will become Market-1 car owners and today’s ride buyer (bus, taxi, carshare, car-renter) will become Market-2 ride buyers. Out of the gate, car-owners remain car owners and ride-buyers remain ride-buyers. Initially, crossover will be miniscule. Given mixed traffic, distracted driving, uncertainty about full automation, and inadequate operating or certification standards for full automation, there will only be normal, organic change in policy volumes and premiums i.e., limited disruption.

But this is a Tortoise and Hare story. In the 2020s, there will be no barriers to Market-1 vehicle purchase and use (except price). The difference is just a better vehicle that should itself be getting safer, but traveling more kilometers on roads that are as effectively complex and unsafe as they are today — perhaps even less safe as increasingly distracted drivers put all proximate travelers at risk thereby sustaining or increasing collision rates, as has been the case in 2015-2016.

While there are no diffusion barriers for Market-1 AVs, there are dozens of barriers for the diffusion of Market-2 ride services: their spatial range is limited, their speed is capped at 40kph, significant investment and set up is needed, social attitudes and behavioural preferences for vehicle ownership, and an initial low network-effect that will restrict range, availability and connectivity, are just a few. Ride-buyers will buy robo-rides as available, but at first very few car-owners will abandon car ownership (and their insurance policy) in favour of using limited robo-taxis. If government is not proactive, competing fleets of commercial robo-taxis forming a service patchwork will not sufficiently encourage car-owners to abandon ownership. The growth of sprawl due to the travel-convenience of Market-1 vehicles will effectively counter any early drop in vehicle ownership due to the initially limited coverage of robo-rides.

Few people will abandon ownership until the on-demand robo-service replacement is robust — i.e., sufficiently reliable, available, and with coverage that matches their travel needs.

Hence, Market-1 vehicles — the Hare in our story — with a human driver to take over as needed to provide 100% trip coverage will diffuse considerably more rapidly than will Market-2 ride-buying. This implies that policy volume will be sustained and likely grow in the near term, even as premiums eventually fall in response to the decline in collisions. At the same time, Market-2 ride services will nibble at the edges of non-automated taxi and transit in ways that will be reported in the press as disruptive of taxi and bus, but would represent statistically insignificant losses for the auto insurance industry — at first. As Market-2 picks up steam it will remain focused on disrupting taxi and bus, even as Market-1 is growing. Projections of 25 or 30 percent penetration of world-wide robotic passenger kilometers traveled by the mid 2030s, will find the majority of their initial customers from the existing legions of taxi, bus and Uber users, having little early effect on absolute levels of car ownership. Hence, the initial impact of AV diffusion is expected to have little effect on policy volumes and premiums.

But this is likely to change in the second decade of AV diffusion. At some point, the range and service levels of shared AV fleets will begin enticing car owners to become ride buyers. More two-car households will become one-car households. Some one-car households will become zero-car households. The diffusion curves will cross and the Tortoise will pass the Hare. This crossover — growing kilometres traveled in shared vehicles that cause vehicle ownership to plateau and decline — is generally projected to occur during the 2030s. While there is no guarantee for this timing, the change will occur. As that happens, the next questions become: how far will ownership fall relative to sharing and how long will it take until it settles into a new plateau? Answers here are far harder to determine.

Complicated by human factors…

In 2032, boomers’ ages will range between 72 and 87. Will these long-time car owners turn decisively to ride buying, or will they prefer fully automated, personal AVs to handle their required travel range for shopping, visiting friends and family, and going to the doctor? This will depend on the quality of local or regional Market-2 services, whether each aging traveler personally judges reliability and security higher in a shared AV with fleet-management oversight or in a private AV with more convenient storage for their assistive devices. When pondering the additional cognitive load of acquiring and using a different vehicle for every trip segment, consider how difficult today’s smartphones are for people over 80.

This example illustrates that for car-owners the shift to non-ownership will often be a complex and personal decision, weighted by several factors, some irrational and many hard to dismiss. Vehicles that consumers purchase are chosen and retained for a spectrum of attributes: convenience, status, privacy, safety, comfort, reliability, immediate availability, job access, or as a marker for career success. Cost per kilometre travelled may be a factor, but it is seldom decisive. If it were, more people would use bicycles or e-bikes.

Human preferences, habits, social biases and the installed infrastructure base that has grown around the private automobile makes the shift away private ownership far more difficult than seen by those projecting a rapid rise in shared vehicle use.

…but follow the money anyway

A switch away from vehicle ownership has many barriers to delay and constrain it, but there are also a number of other forces to negate those barriers. These are largely about money and will become noticeably more menacing for the auto insurance industry post 2030.

City money

Cities strain at every turn to manage current transportation practices and systems: transit, roads, parking, taxi, and enforcement. Two major effects are financial burden and inadequate transit service levels compared to private vehicle usage. This provides a ripe opportunity for Market-2 fleets to disrupt. A preview of this is visible in a 2017 study showing that Uber (a Market-2 surrogate in the 2010s) is adding congestion in NYC and other cities by taking a significant portion of its riders from transit, cycling and walking.

This teaches us that the disruptibility of the current, ride-buying market (including transit as well as taxi) provides the path for Tortoise to gain traction before overtaking Hare. Market-2 can thrive and expand in 2020-2030 without needing to take any car-owners away from Market-1.

The current world of today’s urban surface transportation — with incomprehensibly sub-optimal traffic mixed with struggling transit systems — is the world’s largest market mostly wasted in execution. According to Adam Jonas of Morgan Stanley, this is “… a century-old ecosystem being ogled by outside players hungry for a slice of a US$10- trillion mobility market. Many want in. It’s just beginning. And it won’t stop.” The digital competence of Market-2 will eventually grow to encroach on Market-1 and its insurance counterpart.

New Mobility money

Big Auto and Silicon Valley’s New Mobility ventures see the possibility of monetising more of each consumer’s trip. While digitization and artificial intelligence provide for physical and network optimization as well as greater trip safety, the largest incremental value to its providers is in attention capture — repurposing the traveler’s attention – and her money – from driving to consuming, which includes online shopping and immersive entertainment while inside a moving vehicle. A new focus on selling more value to travelers explains why Big Auto is investing in the ride-selling market.  And that feeds back into early and massive disruption of urban transit systems. The money behind moving existing transit and taxi trips into New Mobility’s digital paradigm is huge. When Tortoise passes Hare he will be a juggernaut, and that threatens the auto insurance industry.

Consumer money

Big Auto wants to be both Hare and Tortoise. Profit from car buyers. Profit from ride buyers. The question really is: can they profit more from selling five or six cars or from renting one car for rides for 12-16 hours per day for 2 years? The truth appears to be the latter — the opportunity for additional services (and optimization) is greater in a Market-2 vehicle. 

Travelers in North America have historically spent some 12 percent of their disposable income on transportation.  More of that can be channelled to other consumption from which ride sellers can now reap an incremental slice — money now left on the table by car sellers, taxi-operators and transit authorities or spent on fuel, parking, wasted assets and labour. The biggest threat to auto insurance is that value capture from digital mobility can turn more car buyers into ride buyers. A growing market share of ride services will be more critical to auto insurance than the fact that car buyers may purchase safer cars.

To the optimizer go the spoils.

 

Further reading

Deloitte
This Deloitte report describes four Future States for the future of mobility that “will co-exist for an extended period of time (perhaps the next 10-15 years)” (p.6). We agree with their description of these states (their States 3 and 4 somewhat parallel our Market-1 and Market-2), but we think the time-line until market maturity and subsequent co-existence will be at least two times longer given the impedance in the systems of automobility that have developed over the past century — at least in the developed world. Deloitte’s recommended responses to these four States by the insurance industry are instructive.

KPMG
This KPMG report describes a rapidly changing technology space that would accelerate and affect the insurance business sooner than many insurance executives appear to believe. We agree with KPMG that the automotive technology will arrive very soon for conditional-automation. However, their projection of an 80% drop in accident frequency by 2040 (p.26) requires a significant rate of diffusion which they did not defend. The vehicular technology is not the barrier, rather deployment infrastructure and social diffusion matter more. We agree with the structural direction and the advice KPMG provides, but caution that their timeline is likely aggressive. Having said that, it is best to anticipate the changes described by KPMG, as they are unstoppable.

Munich RE