Not only are Waymo vehicles driverless, they are often passengerless as well
While Waymo often claims that their vehicles are efficient and clean, they fail to point out a key fact: they are often zero-occupancy vehicles. In fact, for more than 50% of the miles traveled, they are driving around empty. Yes, even worse than the dreaded single-occupancy vehicle that inefficiently clog up our streets.
Once again, we have been given a demonstrably false view that robotaxis will reduce VMT. But it turns out that when they replace car trips, they more than double VMT. And of course when they replace walking, biking, or transit trips then they are adding VMT of twice the length of the trip. This increase in VMT clearly has a negative impact on our cities, by increasing both congestion and pollution.
Now you know why the rear windows of Waymos are tinted. It is so people can’t see that the vehicles are empty more often than not.
Source of Data
Waymo doesn’t directly provide key data since the regulators do not require them to. So that others can verify the data and do more interesting things with it, an explanation of the methodology is key.
The raw data comes from the California PUC AV Program Quarterly Reporting page. The data is provided quarterly towards the bottom of the page. Click on the “Download all reports” link for the desired quarter. This will download an approximately 5MB zip file. Open the zip file and go into the corresponding directory that was created, and then go into the “Waymo” directory, and then into “Driverless Deployment” directory. There are multiple files. Most of them, including the really large one, are almost completely redacted to the public and therefore provide no useful information. This is because Waymo tries as hard as possible to not inform the public.
But there is one really small CSV file that actually contains useful data. It is called something like “PSG0038152_2024_04_AV_Month-Level_Part0.csv” and provides monthly aggregate data for the previous 3 months. It looks like:
While these numbers do not directly include vehicle miles traveled of empty vehicles, that can still fortunately be determined quite accurately. We are provided with the total Vehicle Miles Traveled (TotalVMTZEV). The “ZEV” in the name is for Zero Emission Vehicles, which all Waymos are classified as. One simply need to determine how many miles were traveled with passengers in order to subtract that from the TotalVMTZEV to determine EmptyVMT. And then the percentage of empty (no passengers) miles is simply:
EmptyVMT = TotalVMTZEV – PassengersVMT
and for determining empty percentage:
EmptyPercentVMT = 100% x EmptyVMT / TotalVMTZEV
Let PassengersVMT be the number of miles traveled with at least one passenger inside. This is different from TotalPMT, which is how many passenger miles were traveled in total. To clarify, when a vehicle has 3 passengers and travels 5 miles then it contributes 3×5=15 TotalPMT but only 5 PassengersVMT.
The Waymo data provides TotalTrips, which are number of trips serving at least one passenger. It also provides TotalPassengersCarried. A simple way to estimate PassengersVMT is to use:
AveragePassengersPerTrip = TotalPassengersCarried/TotalTrips
and
PassengersVMT = TotalPMT / AveragePassengersPerTrip
By combining all of the above equations we obtain:
EmptyPercentVMT = 100% x (TotalVMTZEV – (TotalPMT x TotalTrips / TotalPassengersCarried)) /TotalVMTZEV
Time to try out the equation for the last month of available data, February 2024:
EmptyPercentVMT = 100% x (411307 – (258745 x 74233 / 94672)) /411307
EmptyPercentVMT = 50.7% for February 2024
Impressive. For over half the vehicle miles travelled in February, the Waymo vehicles were zero-occupancy vehicles!
For January 2024:
EmptyPercentVMT = 100% x (426808 – (258334 x 77242 / 98240)) /426808
EmptyPercentVMT = 52.3% for January 2024
For December 2023:
EmptyPercentVMT = 100% x (408560 – (262946 x 72595 / 99400)) /408560
EmptyPercentVMT = 52.4% for December 2023
Conclusions
For approximately 52% of the miles traveled, Waymo vehicles are zero-occupancy vehicles, which greatly increases VMT, congestion, and pollution. This means that they are a detriment to our roads, the complete opposite of what the robotaxi companies have been falsely conveying.
Not only is the number of empty vehicle miles traveled appalling, there is 2nd piece of interesting information provided by the data. For 3 months covered in the most recent data, December – February, the monthly ridership actually declined. December was 99,400, January was 98,240, and February was 94,672.
Of course February had only 29 days while January and December had 31 days. The number of riders per day actually went up a small amount from 3,206 in December to 3,264 in February. But that is a paltry increase of just 1.8% .We can at least conclude that usage is not increasing robustly in San Francisco, dashing the hopes of Waymo being able to grow into a viable business.
For Math Pedants
Some sticklers will note that the PassengersVMT (miles traveled with at least one passenger inside) equation used above
PassengersVMT = TotalPMT / AveragePassengersPerTrip
would not be completely accurate if the resulting equations were non-linear. A Waymo with more than a single passenger can actually have 2, 3, or even 4 passengers. Using AveragePassengersPerTrip is a useful simplification. To be thorough it is good to explore the range that PassengersVMT will have depending on how many passengers there are for each trip. For multi-passenger situations assuming there to always be 2 passengers will provide one end of the range, while assuming there to always be 4 passengers will provide the other end of the range. The variance, if any, is likely to be small. But still good to verify.
First we combine
EmptyPercentVMT = 100% x EmptyVMT / TotalVMTZEV
and
EmptyVMT = TotalVMTZEV – PassengersVMT
to get
EmptyPercentVMT = 100% x (TotalVMTZEV – PassengersVMT) / TotalVMTZEV
Next we determine the range of EmptyPercentVMT, based on there always being 2 passengers and 4 passengers. Let N be number of passengers when there is more than a single one. Let:
ExcessPassengers = TotalPassengersCarried – TotalTrips
SinglePassengerTrips = TotalTrips – MultiPassengerTrips
where
MultiPassengerTrips = ExcessPassengers / (N – 1)
We also know that:
PassengersVMT = TotalTrips x AvgMilesPerTrip
We can determine AvgMilesPerTrip via:
(SinglePassengerTrips + N x MultiPassengerTrips) x AvgMilesPerTrip = TotalPMT
or
AvgMilesPerTrip = TotalPMT / (SinglePassengerTrips + N x MultiPassengerTrips)
And using AvgMilesPerTrip in the PassengersVMT equation yields:
PassengersVMT = TotalTrips x TotalPMT / (SinglePassengerTrips + N x MultiPassengerTrips)
where we can use our previous equations to determine EmptyPercentVMT for each N:
ExcessPassengers = TotalPassengersCarried – TotalTrips
MultiPassengerTrips = ExcessPassengers / (N – 1)
SinglePassengerTrips = TotalTrips – MultiPassengerTrips
EmptyPercentVMT for N=2 for February 2024
ExcessPassengers = 94,672 – 74,233 = 20,439
MultiPassengerTrips = 20,439 / (2 – 1) = 20,439
SinglePassengerTrips = 74,233 – 20,439 = 53,794
using PassengersVMT = TotalTrips x TotalPMT / (SinglePassengerTrips + N x MultiPassengerTrips)
PassengersVMT = 74,233 x 258,745 / (53,794 + 2 x 20,439) = 202,884
using EmptyPercentVMT = 100% x (TotalVMTZEV – PassengersVMT) / TotalVMTZEV
EmptyPercentVMT = 100% x (411,307 – 202,884) / 411,307 = 50.67%
EmptyPercentVMT for N=4 for February 2024
ExcessPassengers = 94,672 – 74,233 = 20,439
MultiPassengerTrips = 20,439 / (4 – 1) = 6,813
SinglePassengerTrips = 74,233 – 6,813 = 67,420
using PassengersVMT = TotalTrips x TotalPMT / (SinglePassengerTrips + N x MultiPassengerTrips)
PassengersVMT = 74,233 x 258,745 / (67,420 + 4 x 6,813) = 202,884
202,884 is the same exact value we determined for N=2! This means that the number of passengers for multi-passenger trips actually does not make any difference. If we would go through the equations in detail we would find that N cancels out. So this turned out to be just an exercise in showing that the original calculations were correct.
