My main recommendation this week is that you check out Tim Lee’s very detailed look at Waymo’s safety track record. He does what good skeptical journalists do and really kicks the tires on their claims of safety. In the course of doing so, he finds a couple of instances where Waymo has misclassified or mischaracterized safety incidents. It turns out that they are not quite as good as they claim to be.

But, they’re still pretty damn good!

With all errors corrected, “Waymo’s vehicles crash far less often than human drivers on public roads,” and a majority of the serious safety incidents involving Waymo vehicles occurred because of errors made by other human drivers, which is just to say that Waymo isn’t magic. There are drivers out there running red lights and otherwise misbehaving and crashing into other cars, and some of the cars they crash into are Waymo cars. As the number of self-driving cars on the road goes up, the number of self-driving car crashes will also rise, even if self-driving cars are much safer than human-driven cars, because the human-driven cars are out there crashing into the self-driving ones.

I’ve written a few times about self-driving cars recently, and I covered traffic safety even more recently, and also did a piece about why I think the effort to erect a firewall between the Democratic Party and the private sector is a mistake.

But Tim’s piece brought it all together for me. Waymo is a subsidiary of Alphabet, which is one of the biggest companies in the world. Like all giant companies, Alphabet has a lot of regulatory issues and sometimes the company’s position on those regulatory issues is wrong on the merits, and we need politicians and public servants who can stand up to lobbying money and do the right thing. But the populist conceit that you can ascertain who is correct in a given regulatory battle by just assuming that the big rich company is wrong and its “little guy” opponents are right is not accurate. In the case of self-driving car safety, Waymo is correct and its critics are wrong. Thousands of lives are lost each year to unsafe driving, and life expectance is a major domain in which the United States falls short compared to poorer peer countries in Europe. The fact that Alphabet is right about this doesn’t mean that they are right in every other regulatory issue. But it does mean that making policy judgments based on vibes or a hazy desire to “fight big tech” could do enormous harm.

Just as another example, as part of the American Prospect’s ongoing war against Democrats with private sector experience, in advance of the Trump-Harris debate, they complained that Harris’s debate prep was led by Karen Dunn, the successful lawyer hired by Google for their antitrust showdown with the DOJ. Well, guess what? It turns out that Google is a big rich company that is good at spotting talent and pays top dollar. I don’t have any information about the underlying merits of the antitrust suit, but in my opinion, Karen Dunn is an American hero who laid the groundwork for the best anti-Trump debate performance we’ve seen across three presidential campaigns. It’s good if talented people want to work for you, and sometimes the giant the company is correct in a policy dispute!

Other recommendations:

• This Chad Aldeman data feature on which districts are good at teaching kids to read.

• John Ganz on the smearing of Haitian immigrants.

• Rachel Cohen on factory-build ADUs in California.

This week in good news: Median household income went up a lot in 2023, we could generate tons more electricity out of existing nuclear facilities, transparent mice are cool, this new urban gondola tech is also cool, and a jail in Maine seems to have come up with something constructive on opioids.

Comment of the week from Garrett Cunningham:

It's remarkable how quickly many self-described DC YIMBYs (even ones subscribed to Slow Boring!) retreat when the Height Act is mentioned. "But Paris!" "But the monuments!" "But existing zoning!"

The question is whether being a low-rise city or viewing the monuments (something that can only be done in very specific places anyway) is worth the tradeoffs. How high must our rent prices be so that some rooftops can eke a peek at the Washington Monument? How many tents on the street are worth being low-rise? How high must our taxes be to accommodate a smaller tax base?

Some also point out that DC's Comprehensive Plan doesn't allow for building to the Height Act maximums in many areas, which is true. But locations aren't fungible. The downtown core would be a great place for taller buildings, and other locations are not. These objectors are also often opposed to redeveloping historic neighborhoods (e.g. rowhomes in Logan Circle), though altering the Plan without removing the Height Act would push redevelopment to these historic neighborhoods. You can't have it both ways. Let the market work.

Our question this week was an interesting (and constructive) one from Freddie deBoer, who asks: Why has slowing population growth not attenuated the housing costs problem? Or has it, and we just can't see it because the underlying zoning and regulatory etc situation is just that bad?

I think the basic story here is pretty clear. According to the 2020 Census, “the total number of housing units in the United States grew by 6.7% between 2010 and 2020 — approximately half the rate of growth during the previous decade, according to 2020 Census results released today.” By contrast, the US population grew 7.4 percent that decade, which was a slowdown from 9.7 percent in the previous decade.

This is to say that in the 2000-2010 period, the housing stock grew faster than the population, while in the 2010-2020 period, both growth rates slowed — but the rate of housing stock growth slowed much more. So I think my baseline view is that slowing population growth probably has attenuated housing cost pressures somewhat.

Stepping back, though, it’s worth saying that the relationship between population and housing demand isn’t entirely straightforward. All else being equal, a family of four wants more square footage than a family of two. But it’s not necessarily double the square footage. That’s especially true if you think about households that shrink. You’d expect a middle-aged married couple with two kids to live in a much bigger dwelling than a young couple with no kids. But when that middle aged couples’ children grow up and move out, are they going to downsize back to a much smaller place? They certainly might. Some people do. But lots of people don’t. Nobody finds the prospect of a sixty-something couple with a large suburban home and two empty bedrooms to be particularly puzzling or unusual.

Most American cities fall into one of two buckets.

Many in the South and the West are at their all-time highest population right now. And then most of the ones in the Midwest and some in the Northeast are in a cycle of decline — they peaked around 1950, lost population due to suburbanization, and as of today, continue to shrink and be fairly cheap.

But there’s also a small set of “rebound cities” that loom large in the discourse, largely because Manhattan, Brooklyn, and DC all fall into this category. Rebound cities suffered population crashes in the past, but bottomed out and have rebounded over the past generation without yet re-obtaining their historic population peak. This is really striking in DC, which had 670,000 people in the 2020 Census — which is is a lot more than the 2000 Census low point of 570,000, but a lot less than the 1950 Census high point of 800,000. Given that we are still 130,000 short of peak population and that despite my whining about regulatory constraints there has clearly been a lot of new construction, why is housing still so expensive?

The answer is that DC had many more households in 2020 than it had in 1950, even though it had fewer people. A structure that once upon a time might have held mom, dad, three kids, and a live-in servant is now mom, dad, one kid, a home office, and a TV room.

The shrinking household size is intertwined with rising housing aspirations. It’s very common for families with kids to feel that they have to — or at least want to — move out to someplace where they can afford more space. This seems natural to me, the same as it does to everyone else. But factually speaking, back in the peak Baby Boom years, when people had larger families and the population growth rate was much faster, the houses were smaller. The original Levittown houses were 750 square feet, with a larger 1,000 square feet option also available. The average newly built home around then was 1,200 square feet, and of course, large numbers of people lived in old dwellings that were often smaller.

This means you could make a lot of different linkages between housing supply, housing demand, and population size:

• If people had fewer children, there would be less aggregate demand for square footage and housing would be cheaper.

• If housing supply were less constrained, square footage would be cheaper and people would have more children.

• If people reduced their demand for square footage, they would feel comfortable in smaller homes and would have more kids.

• If people had more kids, they would be more comfortable with crowding and the number of people per square foot would go up.

All of these ideas are certainly plausible. The fact is that preferences and norms and cultural expectations are a little murky. But one thing we know is that if you look at very expensive cities like Los Angeles, what tends to happen is that native-born Americans leave, because they want to go someplace where they can afford a bigger house. People from other countries, meanwhile, are often accustomed to more crowding back home, so they find the tradeoff in favor of living in the expensive city to be more appealing.

All of this (gestures vaguely) is one reason why I personally don’t love talk about “affordability” or how many units a country or city or state “needs.”

Instead, in a habit that I imagine Freddie finds appalling, I like to take a really simplistic Economics 101 view. It is important to have regulations to avoid pollution externalities. There is also a risk that due to asymmetrical information, buildings will be unsafe unless there are rules in place. But in general, if someone wants to build something, they should be allowed to do that. When you put rules in place that block them from doing so, rules that are not motivated by bona fide environmental externalities or public safety concerns — rules limiting lot size or building height, rules requiring parking, rules limiting Floor Area Ratio, rules requiring safety features with no value, rules banning multi-family construction from certain areas — the result is a lot of economic loss and inefficiency. It’s challenging to predict in advance what exactly the nature of those losses will be or what exactly would happen in a world of housing freedom. Would houses stay the same and per household expenditures fall? Would expenditures stay flat and homes get much larger? A little of both? Maybe since the price per square foot is lower, people would consume so much more housing (more vacation houses, for example) that expenditures go up? Would the population growth rate increase a lot because people have more kids? Would the public become more pro-immigrant without concerns about the housing impacts?

I have guesses, and there are ways to gather evidence, but it is fundamentally hard to predict in advance beyond “a huge source of inefficiency will go away.”

It’s Autonomy Week! This is the second of five articles exploring the state of the self-driving industry.

As a reminder, annual subscriptions to Understanding AI are $63.75, a 15 percent discount over the regular annual price. The discount is available until Friday.

On a Friday evening last November, police chased a silver sedan across the San Francisco Bay Bridge. The fleeing vehicle entered San Francisco and went careening through the city’s crowded streets. At the intersection of 11th and Folsom streets, it sideswiped the fronts of two other vehicles, veered onto a sidewalk, and hit two pedestrians.

According to a local news story, both pedestrians were taken to the hospital with one suffering major injuries. The driver of the silver sedan was injured, as was a passenger in one of the other vehicles.

No one was injured in the third car, a driverless Waymo robotaxi. Still, Waymo was required to report the crash to government agencies. It was one of 20 crashes with injuries that Waymo has reported through June.  And it’s the only crash Waymo has classified as causing a serious injury.

Twenty injuries might sound like a lot, but Waymo’s driverless cars have traveled more than 22 million miles. So driverless Waymo taxis have been involved in fewer than one injury-causing crash for every million miles of driving—a much better rate than a typical human driver.

On Thursday, Waymo released a new website to help the public put statistics like this in perspective. Waymo estimates that typical drivers in San Francisco and Phoenix—Waymo’s two biggest markets—would have caused 64 crashes over those 22 million miles. So Waymo vehicles get into injury-causing crashes less than one-third as often, per mile, as human-driven vehicles.

Waymo claims an even more dramatic improvement for crashes serious enough to trigger an airbag. Driverless Waymos have experienced just five crashes like that, and Waymo estimates that typical human drivers in Phoenix and San Francisco would have experienced 31 airbag crashes over 22 million miles. That implies driverless Waymos are one-sixth as likely as human drivers to experience this type of crash.

The new data comes at a critical time for Waymo, which is rapidly scaling up its robotaxi service. A year ago, Waymo was providing 10,000 rides per week. Last month, Waymo announced it was providing 100,000 rides per week. We can expect more growth in the coming months.

So it really matters whether Waymo is making our roads safer or more dangerous. And all the evidence so far suggests that it’s making them safer.

It’s not just the small number of crashes Waymo vehicles experience—it’s also the nature of those crashes. Out of the 23 most serious Waymo crashes, 16 involved a human driver rear-ending a Waymo. Three others involved a human-driven car running a red light before hitting a Waymo. There were no serious crashes where a Waymo ran a red light, rear-ended another car, or engaged in other clear-cut misbehavior.

In total, Waymo has reported nearly 200 crashes, which works out to about one crash every 100,000 miles. Waymo says 43 percent of crashes across San Francisco and Phoenix had a delta-V of less than 1 mph—in other words, they were very minor fender-benders.

But let’s focus on the 23 most severe crashes: those that either caused an injury, caused an airbag to deploy, or both. These are good crashes to focus on not only because they do the most damage, but because human drivers are more likely to report these types of crashes, making it easier to compare Waymo’s software to human drivers.

A large majority of these—16 crashes in total—involved another car rear-ending a Waymo. Some were quite severe: three triggered airbag deployments and one caused a “moderate” injury. One vehicle rammed the Waymo a second time as it fled the scene, prompting Waymo to  sue the driver.

There were three crashes where a human-driven car ran a red light before crashing into a Waymo:

• One was the crash I mentioned at the top of this article. A car fleeing the police ran a red light and slammed into a Waymo, another car, and two pedestrians, causing several injuries.

• In San Francisco, a pair of robbery suspects fleeing police in a stolen car ran a red light “at a high rate of speed” and slammed into the driver’s side door of a Waymo, triggering an airbag. The suspects were uninjured and fled on foot. The Waymo was thankfully empty.

• In Phoenix, a car ran a red light and then “made contact with the SUV in front of the Waymo AV, and both of the other vehicles spun.” The Waymo vehicle was hit in the process, and someone in one of the other vehicles suffered an injury Waymo described as minor.

There were two crashes where a Waymo got sideswiped by a vehicle in an adjacent lane:

• In San Francisco, Waymo was stopped at a stop sign in the right lane when another car hit the Waymo while passing it on the left.

• In Tempe, Arizona, an SUV “overtook the Waymo AV on the left,” then “initiated a right turn,” cutting the Waymo off and causing a crash. A passenger in the SUV said they suffered moderate injuries.

Finally, there were two crashes where another vehicle turned left across the path of a Waymo vehicle:

• In San Francisco, a Waymo and a large truck were approaching an intersection from opposite directions when a bicycle behind the truck made a sudden left in front of the Waymo. Waymo says the truck blocked Waymo’s vehicle from seeing the bicycle until the last second. The Waymo slammed on its brakes but wasn’t able to stop in time. The San Francisco Fire Department told local media that the bicyclist suffered only minor injuries and was able to leave the scene on their own.

• A Waymo in Phoenix was traveling in the right lane. A row of stopped cars was in the lane to its left. As Waymo approached an intersection, a car coming from the opposite direction made a left turn through a gap in the row of stopped cars. Again, Waymo says the row of stopped cars blocked it from seeing the turning car until it was too late. A passenger in the turning vehicle reported minor injuries.

It’s conceivable that Waymo was at fault in these last two cases—it’s impossible to say without more details. It’s also possible that erratic braking by Waymo contributed to a few of those rear-end crashes. Still, it seems clear that a non-Waymo vehicle bore primary responsibility for most, and possibly all, of these crashes.

One should always be skeptical when a company publishes a self-congratulatory report about its own safety record. So I called Noah Goodall, a civil engineer with many years of experience studying roadway safety, to see what he made of Waymo’s analysis.

“They've been the best of the companies doing this,” Goodall told me. He noted that Waymo has a team of full-time safety researchers that publishes their work in reputable journals.

Waymo knows exactly how often its own vehicles crash because its vehicles are bristling with sensors. The harder problem is calculating an appropriate baseline for human-caused crashes.

That’s partly because human drivers don’t always report their own crashes to the police, insurance companies, or anyone else. But it’s also because crash rates differ from one area to another. For example, there are far more crashes per mile in downtown San Francisco than in the suburbs of Phoenix.

Waymo tried to account for these factors as it calculated crash rates for human drivers in both Phoenix and San Francisco. To ensure an apples-to-apples comparison, Waymo’s analysis excludes freeway crashes from its human-driven benchmark, since Waymo’s commercial fleet doesn’t use freeways yet.

Waymo estimates that human drivers fail to report 32 percent of injury crashes; the company raised its benchmark for human crashes to account for that. But even without this under-reporting adjustment, Waymo’s injury crash rate would still be roughly 60 percent below that of human drivers. The true number is probably somewhere between the adjusted number (70 percent fewer crashes) and the unadjusted one (60 percent fewer crashes). It’s an impressive figure either way.

Waymo says it doesn’t apply an under-reporting adjustment to its human benchmark for airbag crashes, since humans almost always report crashes that are severe enough to trigger an airbag. So it’s easier to take Waymo’s figure here—an 84 percent decline in airbag crashes—at face value.

Waymo’s benchmarks for human drivers are “about as good as you can do,” Goodall told me. “It's very hard to get this kind of data.”

When I’ve talked to other safety experts, they’ve been equally positive about the quality of Waymo’s analysis. For example, last year I asked Phil Koopman, a professor of computer engineering at Carnegie Mellon, about a previous Waymo study that used insurance data to show its cars were significantly safer than human drivers. Koopman told me Waymo’s findings were statistically credible, with some minor caveats.

Similarly, David Zuby, the chief research officer at the Insurance Institute for Highway Safety, had mostly positive things to say about a December study analyzing Waymo’s first 7.1 million miles of driverless operations.

If you look closely, you’ll see that one of the numbers in this article differs slightly from Waymo’s safety website. Specifically, Waymo says that its vehicles get into injury crashes 73 percent less often than human drivers, while the figure I use in this article is 70 percent.

This is because I spotted a couple of apparent classification mistakes in the raw data Waymo used to generate its statistics.

Each time Waymo reports a crash to NHTSA, it records the severity of injuries caused by the crash. This can be fatal, serious, moderate, minor, none, or unknown.

When Waymo shared an embargoed copy of its numbers with me early last week, it said that there had been 16 injury crashes. But when I looked at the data Waymo had submitted to federal regulators, it showed 15 minor injuries, two moderate injuries, and one serious injury, for a total of 18.

When I asked Waymo about this, the company said it found a programming error. Waymo had recently started using the “moderate” injury category, and had not updated the code that generated its crash statistics to count these crashes. Waymo fixed the error quickly enough that the published version of Thursday’s report showed 18 injury crashes.

But as I continued looking at the data, I noticed another apparent mistake: two crashes had been put in the “unknown” injury category, yet the narrative for each crash indicated an injury had occurred. One report said “the passenger in the Waymo AV reported an unspecified injury.” The other stated that “an individual involved was transported from the scene to a hospital for medical treatment.”

I notified Waymo about this apparent mistake on Friday and they say they are looking into it. As I write this, the website still claims a 73 percent reduction in injury crashes. But I think it’s clear that these two “unknown” crashes were actually injury crashes. So all of the statistics in this article are based on the full list of 20 injury crashes.

I think this illustrates that I come by my generally positive outlook on Waymo honestly: I probably scrutinize Waymo’s data releases more carefully than any other journalist, and I’m not afraid to point out when the numbers don’t add up.

Based on my conversations with Waymo, I’m convinced these were honest mistakes rather than deliberate efforts to cover up crashes. I was only able to identify these mistakes because Waymo went out of its way to make its findings reproducible. It would make no sense to do that if they were simultaneously trying to fake their own statistics.

Could there be other injury or airbag-triggering crashes that Waymo isn’t counting? It’s certainly possible, but I doubt there have been very many. You might have noticed that I linked to local media reporting for some of Waymo’s most significant crashes. If Waymo deliberately covered up a serious crash, there’d be a big risk that a crash would get reported in the media and then Waymo would have to explain to federal regulators why it wasn’t reporting all legally required crashes.

So despite the screwups, I find Waymo’s data to be fairly credible, and those data show that Waymo’s vehicles crash far less often than human drivers on public roads.

Support from paid subscribers allows me to do the kind of in-depth reporting you find in this article. So if you enjoyed today’s story, please upgrade. You’ll get access to subscriber-only articles and enable me to continue doing in-depth journalism about the AI industry.

This week annual subscriptions are just $63.75, a 15 percent discount over the regular annual rate.