Tesla Releases Q2 Safety Report, Adds Vehicle Fire Stats

Tesla released this week its Q2 2019 vehicle safety report, which for the first time included stats for vehicle fires. 

“Tesla vehicle fires are exceptionally rare events, and in some cases, there have been zero Tesla vehicle fires in a quarter,” the report said. However, an increase from one fire per quarter to two per quarter represents an increase of 100%, so in order to avoid misinterpretation of the numbers and provide a meaningful comparison to industry data, Tesla says it will publish an update to vehicle fire data annually.

From 2012 – 2018, there has been approximately one Tesla vehicle fire for every 170 million miles traveled, Tesla reported. By comparison, data from the National Fire Protection Association (NFPA) and U.S. Department of Transportation shows that in the U.S. there is a vehicle fire for every 19 million miles traveled.

In order to provide an apt comparison to NFPA data, Tesla’s data set includes instances of vehicle fires caused by structure fires, arson, and other things unrelated to the vehicle, which account for about 15% of Tesla vehicle fires over this time period.

The safety report also showed that Teslas operating on Autopilot are less likely to get into accidents compared to vehicles not using a self-driving system. 

In Q2, Tesla registered one accident for every 3.27 million miles driven in which drivers had Autopilot engaged. For those driving without Autopilot but with active safety features, Tesla registered one accident for every 2.19 million miles driven. For those driving without Autopilot and without active safety features, it registered one accident for every 1.41 million miles driven. For comparison, NHTSA’s data currently stands at one accident for every 498,000 miles driven.

 

[one true morty]

Our middle son's Jeep burst into flames spontaneously while parked in the driveway. Our office manager's T-bird did the same parked at work one day. My perspective is that it is not unusual for gasoline powered cars to catch fire!

My Model 3 sensed me possibly rear- ending the car in front of me in heavy traffic on the 101 and both warned me and began breaking. I probably would have avoided an accident on my own, but was greatful for the automatic response!

A jeep spontaneously burst into flames in front of my house a ways back. That was a terrifying way to wake up. I use autodrive on trips all of the time and my car has helped me avoid a number of sticky situations based on other drivers. I always remain vigilant but it is a relief having the car have my back (or vise versa). I still require it to obtain permission to change lanes however. After reading a few of the accounts of not seeing large trucks and trying to pull under them I just need the extra protection for now.

 

[ItsNotAboutTheMoney]

Yes, so it is hard to understand the thinking behind publishing the data.

The thinking is probably that every time there's a dumb Tesla article, the dumb journalist can easily get the dumb statistics.

 

[Daniel in SD]

Your answer to your first question lies in your second one. Maybe Tesla drivers aren't better drivers, but Teslas are safer cars and more modern cars, their brakes and tyres amongst others are usually well kept.

Well if these numbers are true I'm definitely getting ripped off on my liability insurance. Hopefully Tesla's liability insurance will be a quarter the price

 

[AlanSubie4Life]

Well if these numbers are true I'm definitely getting ripped off on my liability insurance. Hopefully Tesla's liability insurance will be a quarter the price

Yeah, my insurance prices definitely do not reflect the alleged 5x safety improvement, particularly for liability. I really need to look at another company this renewal. Such a pain.

It seems like pretty much everyone agrees these numbers are meaningless as presented. It’s too bad, I don’t understand why a smart guy like Musk continues to act like a salesman with this kind of stuff. Just present the real story - I’m fairly sure it won’t be a BAD story. Just not as good as what they suggest.

 

[nthn]

If you guys all think that the higher numbers are due to safer environments with Autopilot on, road conditions (confounding variables), just cut the results in half from 3.27 mm, to 1.64 mm. Still 3 times better than ICE vehicles. You cannot attribute 6 times better safety record to just confounding variables.

Also, I saw one person said that the numbers didn't include accidents that where avoided. Wasn't that the whole point? The ICE stats don't include that either, so it is fair there.

 

[jsmay311]

“From 2012 – 2018, there has been approximately one Tesla vehicle fire for every 170 million miles traveled, Tesla reported. By comparison, data from the National Fire Protection Association (NFPA) and U.S. Department of Transportation shows that in the U.S. there is a vehicle fire for every 19 million miles traveled.”

170 million miles vs 19 million miles. So nine times better.

 

[AlanSubie4Life]

just cut the results in half from 3.27 mm, to 1.64 mm.

Where did this factor of two come from?

Since I already have a Tesla, I would like to know whether it is safer or not for me to use Autopilot on a given drive. That is the relevant statistic for me. Currently it is not clear what the answer is to that.

It’s quite likely that Teslas get in fewer accidents than a much older fleet of cars, when this older fleet is driving: on undivided highways, in inclement weather, on city streets, with bald tires, etc. But why doesn’t Tesla just try to control for at least some of these factors, so we know what the safety improvement actually is? Since all the data is known by Tesla, I would prefer to not just make up the results.

 

[nthn]

Where did this factor of two come from?

Since I already have a Tesla, I would like to know whether it is safer or not for me to use Autopilot on a given drive. That is the relevant statistic for me. Currently it is not clear what the answer is to that.

It’s quite likely that Teslas get in fewer accidents than a much older fleet of cars, when this older fleet is driving: on undivided highways, in inclement weather, on city streets, with bald tires, etc. But why doesn’t Tesla just try to control for at least some of these factors, so we know what the safety improvement actually is? Since all the data is known by Tesla, I would prefer to not just make up the results.

What I'm saying is, no amount of confounding variables can account for a 6 times increase. So even if you say half of the increase is due to other variables (newer vehicle, safer driving conditions etc.), you STILL have a 3 times increase in safety.

So while I agree with you that more info is better, there is NO way that anyone can say Tesla's are not safer.

 

[Daniel in SD]

What I'm saying is, no amount of confounding variables can account for a 6 times increase. So even if you say half of the increase is due to other variables (newer vehicle, safer driving conditions etc.), you STILL have a 3 times increase in safety.

So while I agree with you that more info is better, there is NO way that anyone can say Tesla's are not safer.

We have no idea how Tesla is defining accidents! It’s obvious they’re not using the same definition as the NHTSA. Perhaps they’re only counting airbag deployments...
The question isn’t whether Teslas are safer than the average car, they certainly are.

 

[Wol747]

Also, I saw one person said that the numbers didn't include accidents that where avoided. Wasn't that the whole point? The ICE stats don't include that either, so it is fair there.

It will all come out in the wash if FSD ever comes about because there would be cars without manual controls so apples will be compared with apples and not oranges.

At present the most significant figure has to be the number of interventions judged to be justified - and of course in the determination of THAT hindsight is a wonderful thing!

 

[AlanSubie4Life]

What I'm saying is, no amount of confounding variables can account for a 6 times increase. So even if you say half of the increase is due to other variables (newer vehicle, safer driving conditions etc.), you STILL have a 3 times increase in safety.

So while I agree with you that more info is better, there is NO way that anyone can say Tesla's are not safer.

I think it is very likely that Telsas are safer than the average car.

Here are some things this data does NOT tell us:

1) Is using Autopilot safer than not using Autopilot, for a given drive?

2) Do Teslas have lower accident rates than other new cars? How about when using Autopilot as compared to other cars on the same roads?

3) Is Tesla safety improving with time? Has accident severity been reduced, if not the frequency (in this data set the changes are very small)?

4) Are the Tesla “non-autopilot” miles from a group of Tesla cars that match the autopilot miles in terms of age/type/safety features?

5) What is the definition of an accident? Do fender benders count? Is it the same definition as NHTSA uses?

The irony here is that Tesla would probably get a lot more free press if they published data that could be examined. Right now it is ignored because even journalists can see it is fairly meaningless data. But the truth is, it’s likely a good story for Tesla, and possibly even for advanced driver assistance systems.

 

[Daniel in SD]

I think it is very likely that Telsas are safer than the average car.

Here are some things this data does NOT tell us:

1) Is using Autopilot safer than not using Autopilot, for a given drive?

2) Do Teslas have lower accident rates than other new cars? How about when using Autopilot as compared to other cars on the same roads?

3) Is Tesla safety improving with time? Has accident severity been reduced, if not the frequency (in this data set the changes are very small)?

4) Are the Tesla “non-autopilot” miles from a group of cars that match the autopilot miles in terms of age/type/safety features?

5) What is the definition of an accident? Do fender benders count? Is it the same definition as NHTSA uses?

The irony here is that Tesla would probably get a lot more free press if they published data that could be examined. It’s likely a good story for Tesla, and possibly even for advanced driver assistance systems.

6) Do Teslas catch on fire less than other vehicles of the same age? (not that it really matters since fires are so rare)

 

[AlanSubie4Life]

What I'm saying is, no amount of confounding variables can account for a 6 times increase.

In general, this is a false statement. The fact that the US fleet average is 6 times less safe than Teslas (or whatever the factor is) implies that there are some extremely unsafe vehicles and situations out there! The average is 6 times worse, so there are probably situations which are 20-30 times less safe! So somewhat tautologically, the confounding factors obviously DO play a huge role. A factor of 6 would be no big deal to achieve when comparing sufficiently dissimilar situations.

That being said, Teslas are very likely safer than the average car.

But the safety of Telsas and Autopilot relative to other new vehicles with safety assistance features (which is the most relevant comparison when choosing, on the basis of safety, between new vehicles!) is likely going to be a much closer call. It may well still be a positive story for Tesla. No one knows but them!

 

[SVMike]

Yes, so it is hard to understand the thinking behind publishing the data.

The most interesting part of the data is the apparent miles/accident “improvement” over the last two quarters, which one might think is due to active safety improvements. But I started thinking about it, and it looks like most of this is seasonal - hard to say exactly how much until we have a couple years of data. We can try to figure out the seasonal component though.

It is much more likely that the last two quarters show an improvement because the vast bulk of the cars being considered are brand new model 3s! I don't know how much safer a less than one year old car is than, perhaps a 2-4 year old car. But the last 2 quarters of 2018 is when Tesla began selling large numbers of cars, far over shadowing the "older" Model S/X.
We are looking at the tip of the iceberg in terms of data and we need several years of data, at least, with lots of cars in lots of locations.

Another factor. Like it or not most Tesla's are sold in CA where the weather is not as big of a factor. The data needs to be compared annually in the same locations against ICE cars. Again, the dataset is too small still.

But it is important to start the rigorous data collection now so we can see what happens when there are more Tesla cars, cheaper cars, more cars in snow/ice, etc.

 

[wschwisow]

Determining that would require hiring a good statistician.
I find it highly unlikely that the accident rate of Teslas is 5 times better than the average car. Even if Tesla drivers are the best drivers on earth we still get run into. These stats seem ridiculous.
Also, comparing fire statistics of cars with an average fleet age of about one to two years to ICE vehicles with an average age of 12 years is pretty misleading as well.

Keep in mind the stats are for AP, which currently means freeway only driving. Even with the navigate on auto option, the stats for freeway accidents will always be much lower than comparing city driving accident rates, so it's really apples to oranges at this point. The stats on Tesla fires said from 2012 to current, not just the latest M3 batch. Is that what you were referring to with fleet age? I would guess that one more accurate (or less misleading) than the accidents per mile stat.

 

[Daniel in SD]

Keep in mind the stats are for AP, which currently means freeway only driving. Even with the navigate on auto option, the stats for freeway accidents will always be much lower than comparing city driving accident rates, so it's really apples to oranges at this point. The stats on Tesla fires said from 2012 to current, not just the latest M3 batch. Is that what you were referring to with fleet age? I would guess that one more accurate (or less misleading) than the accidents per mile stat.

Do they say it's freeway only driving?
Tesla has sold more cars in the last 6 quarters than they did in all the quarters before that. The average age of their cars is about a year. The average age of all registered vehicles in the US is 12 years. If brand new cars catch on fire less often then the stat is misleading.

 

[wschwisow]

Well they say it's auto pilot driving, which is freeway driving until full self driving becomes reality one day. So my point is that they aren't comparing Tesla AP accident rates with other "freeway only" accident rates. So Tesla rates look deceivingly good against other vehicles, which include city driving where more accidents take place. It's probably the only comparison Tesla can make at the moment, but still apples to oranges.

 

[AlanSubie4Life]

It's probably the only comparison Tesla can make at the moment, but still apples to oranges.

Basically agree with you that the data is deceiving, but I'm fairly sure Tesla could make a much more accurate comparison if they so desired; their data is probably very granular. I don't think they are limited to the comparison they've made. And that's a shame.

 

[Daniel in SD]

Well they say it's auto pilot driving, which is freeway driving until full self driving becomes reality one day. So my point is that they aren't comparing Tesla AP accident rates with other "freeway only" accident rates. So Tesla rates look deceivingly good against other vehicles, which include city driving where more accidents take place. It's probably the only comparison Tesla can make at the moment, but still apples to oranges.

I agree but why is the non autopilot accident rate so much better than other cars? Even if Tesla drivers got in to zero at fault accidents they would still need to avoid an unrealistically high percentage of accidents where the other driver would be at fault.

 

[AlanSubie4Life]

I agree but why is the non autopilot accident rate so much better than other cars?

It's due to the incredible safety & prescient nature of Tesla cars.

Seriously though, I'm pretty sure you already answered this question yourself above:

1) Different accident definition
2) Different area of the country (seems like this would bias things the wrong way though)
3) Different driver risk factors
...etc.

There are a lot of possibilities!