How autonomous vehicles will change everything
Speculative predictions on car insurance, real estate, car designs and more
With everyone focused on AI, there’s been surprisingly little discussion on the fact that autonomous vehicles are finally here. Waymo is scaling rapidly (50k weekly trips in May, 100k in August and 150k in October) and surpassed Lyft’s market share in SF, according to Yipit. Tesla Full-Self-Driving Mode (FSD) is improving exponentially, both based on the real data and my small anecdata (aka my one friend who has a Tesla).
So I decided to focus this month’s Charts and Tarts1 on Autonomous Vehicles (hereafter abbreviated to “AVs”) but with a twist. Instead of focusing on “who will win”, I think it’s more interesting to dig into the second-order impacts of AVs on our lives. How will AVs impact insurance costs? Real estate? Car designs?
Below are some charts shared at the event and a summary of the discussions we had.
AVs will drastically reduce insurance costs
Safety is the biggest appeal of AVs. Over 40k Americans are killed every year by car crashes. We have yet to see a Waymo fatality and it’s hard to find an example of a serious accident where Waymo was at fault (as opposed to it getting rear-ended for example).
The data supports this narrative that AVs are safer than human drivers. The chart above from a study that looked at insurance claims data found that Waymo's AVs achieved a 92–100% reduction in bodily injury claims and a 76–95% reduction in property damage claims, depending on the operational mode, compared to human drivers.2
The most interesting debates we had on this were whether we’ll have huge insurance differences in the future based on whether you are using a personal AV or not (and how much that can offset some of the extra hardware/software costs for the AV). Could regular car insurance be $200 a month but AV car insurance be $20 a month?3 Car insurance is a de-facto tax on the cost of operating a car, so this should put savings back into consumers’ pockets.
As for insurance companies, I’d expect them to benefit in the short term given lower than expected claims as AVs proliferate. But we may also have a much smaller industry (in $ of premiums) in the future as a result.
Repricing far away real estate
Most cities see a standard fall in real estate prices (per square foot) as you get further away from the city center (see chart). AVs are likely to flatten this curve, increasing the relative price per square foot of the exurbs/suburbs compared to the city center.
Why? If commuting time becomes productive or comfortable (you can work or sleep on the ride), living further from the city center becomes much more attractive. A 1-hour commute by AV is much much better than driving 1 hour yourself.
In the New York metro where I live, I’d expect two types of areas to benefit:
Places within a commutable distance but have no public transportation. Plenty of towns in New Jersey, Long Island, and Westchester that have <1 hour commutes today by car but have bad public transportation. Driving from these areas today into NYC is pretty miserable.
The exurbs that are too far to reliably commute today but become palatable in the future. Places like Sussex or Putnam counties. You could imagine places like Westchester (cusp of commuting zone, good public transit) actually lose out on a relative basis.
COVID pushed up real estate prices in the suburbs more than the city center as people could now work from home more. This is going to have the same effect, but on steroids.
A traffic and parking nightmare (without congestion pricing)
If everyone wants to drive their personal AVs to work, you can immediately see the big problem:
Much more traffic in the short term4
Much more parking needed
There are a few potential solutions that come to mind, but none of them are satisfying. A shared fleet of vehicles like Waymo may actually increase total vehicle miles, and therefore traffic. That’s because, unlike privately owned cars, these Waymos will sometimes drive without a passenger to reposition themselves (deadhead miles). AVs are also likely to take market share from public transportation, further adding miles to the system. This is why studies have shown Uber and Lyft generally increase total vehicle miles traveled.
If a high share of AV rides are shared between people (similar to UberPool) then it at least has a chance of reducing traffic. But our experience with Uber and Lyft suggests people aren’t willing to share rides at high enough rates because the cost savings aren’t worth the hassle associated with it. I’d expect this to be no different with AVs.5
As for parking, AVs could help with that but at the expense of traffic. The more people use Waymo, the less parking we need (but the worse the traffic problem becomes for the reasons mentioned above). Personal AVs could also drop you off at your workplace and go to some other “parking” zone (presumably further from downtown), but this similarly adds lots of vehicle miles into the system and I doubt people would do this if it risks not having immediate access to your vehicle.
Unless we can build lots of new roads, NYC-style congestion pricing seems like the beginning of what’s to come if we want to avoid big traffic increases.6 Expect much higher parking rates in the future too.
New creativity in car design
Most cars today look remarkably similar. Other than the Cybertruck, it’s hard to remember a recent car that’s looked truly “different.”
One participant predicted a big shift in car design as autonomous vehicles proliferate:
The lack of a need for a driver means you can create more interesting car layouts. Unlike Waymo, which repurposed some Jaguars, Zoox has a truly differentiated design for its autonomous car with seats facing inward and open sliding doors.
Once cars drive themselves, people will spend far more hours in their vehicles—commuting, relaxing, working, sleeping. If you’re in a car for longer stretches, you might want a space tailored to your comfort or lifestyle. That could lead to more customization and a boom in creative vehicle interiors, rather than a bland one-size-fits-all approach. Maybe you’ll have a workstation in your car or a bed.
The main challenge to this is regulation and standardization. Apparently one of the reasons all cars look the same is due to lesser-known regulations and you can imagine all the safety risks these new designs could create. But these are solvable problems, so I expect them to be fixed.
Other predictions from the group
There were a few other interesting predictions from attendees:
Segregated lanes for autonomous vehicles: It will take a while for roads to be 100% AVs, which would maximize speeds, but in the meantime we are likely to have completely segregated lanes on highways where AVs can coordinate with each other and travel marginally faster than today.
“Right to drive” movement: Given the safety risks from human drivers, there will be political pressure to ban humans from being able to drive in most, if not all, situations. We might therefore see a “Right to Drive” movement pop up in America which emphasizes personal freedoms and our G-d given right to drive our cars when we want and how we want!
Some crimes become easier with AVs, other crimes harder: Without a human driver to deter wrongdoing, you could imagine criminals simply putting a cone on an AV (or something similar) to rob passengers. Stealing cars becomes much harder though, as AVs can’t be driven away and are closely tracked.
Waymo will license its technology to automakers: This comes from a great podcast with Derek Thompson and Timothy Lee where Lee predicts Waymo will 1) license its tech to automakers and 2) provide “self-driving” to any car equipped with its tech in any service area it already operates. So your car would have “self-driving mode” in San Francisco but not in New York City (until Waymo launches coverage there).
To get more into the details: The chart compares liability claims for bodily injury and property damage between human-driven vehicles (baseline) and Waymo operations, categorized into three modes:
Rider-Only (RO): Fully autonomous mode with no human driver present.
Testing Operations (TO): Vehicles operated autonomously under the supervision of a human safety specialist.
TO+RO: Combined data from both RO and TO modes to evaluate overall performance.
The "TO+RO" aggregation provides a comprehensive view of Waymo's safety performance across all autonomous miles driven. Each claim frequency is expressed as claims per million miles (cpmm), and statistical significance is indicated by non-overlapping confidence intervals.
Of course, even if AVs are not getting into accidents, insurance will need to cover when the AV (and passenger) are hit by dangerous human drivers still on the roads.
Traffic could be reduced long term if all cars were AVs. Most people don’t realize that traffic today is largely caused by human error, as opposed to accidents or construction. One person slowing down slightly can cause a large traffic build-up a few minutes later.
Cliff Sossin has a fascinating argument why AVs will increase personal car ownership, not decrease it. I may write about this in a separate piece given it's too much to cover here.
Without congestion pricing, even building more roads is unlikely to materially reduce traffic for all the induced demand reasons you typically read about.