I don't think the biggest issue with the LendingTree data is random error, but instead systemic error.
A minor accident is unlikely to appear on the carfax of a 20 year old Pontiac (Pontiac stopped making cars many years ago so they are all old) because who reports a minor accident on a $1000 car?
A minor accident on a newish $50k car is likely to be reported, as it is likely to go through insurance. Tesla has been growing faster than other car brands, the vast majority of their cars are still under 5 years of age, unlike basically every brand of car.
The fact that the report consistently listed brands that no longer make cars as "safe" and new car brands as "unsafe" I think shows that there is likely little random error, but the systemic error dominates the data.
Yeah, that’s a good example of why this data doesn’t make any sense. I tried to emphasize the fundamental problems in this approach, but the biggest challenge in criticizing it is that we don’t even really know what they did to produce these numbers. There doesn’t seem to be an explanation that would justify their claims, but it would really be helpful to know what they did, to be able to more precisely explain just how misleading they are.
Thank you for this great post. Indeed it is not Tesla drivers that had the highest rate of accidents, rather it was the people who got a quote for a Tesla car insurance on LendingTree. These people are looking to buy a Tesla car or renew their existing Tesla insurance. And the accident data is about their self reported past driving history. For example, someone who had BMW, Ford and Chevy cars in the past but now asked for a insurance quote for a Tesla on LendingTree is categorized as Tesla driver. He had an accident while using a BMW for instance, but his accident is counted under Tesla in this stupid study. You can't make this up. What a terrible research design!
I don't think the biggest issue with the LendingTree data is random error, but instead systemic error.
A minor accident is unlikely to appear on the carfax of a 20 year old Pontiac (Pontiac stopped making cars many years ago so they are all old) because who reports a minor accident on a $1000 car?
A minor accident on a newish $50k car is likely to be reported, as it is likely to go through insurance. Tesla has been growing faster than other car brands, the vast majority of their cars are still under 5 years of age, unlike basically every brand of car.
The fact that the report consistently listed brands that no longer make cars as "safe" and new car brands as "unsafe" I think shows that there is likely little random error, but the systemic error dominates the data.
Yeah, that’s a good example of why this data doesn’t make any sense. I tried to emphasize the fundamental problems in this approach, but the biggest challenge in criticizing it is that we don’t even really know what they did to produce these numbers. There doesn’t seem to be an explanation that would justify their claims, but it would really be helpful to know what they did, to be able to more precisely explain just how misleading they are.
Thank you for this great post. Indeed it is not Tesla drivers that had the highest rate of accidents, rather it was the people who got a quote for a Tesla car insurance on LendingTree. These people are looking to buy a Tesla car or renew their existing Tesla insurance. And the accident data is about their self reported past driving history. For example, someone who had BMW, Ford and Chevy cars in the past but now asked for a insurance quote for a Tesla on LendingTree is categorized as Tesla driver. He had an accident while using a BMW for instance, but his accident is counted under Tesla in this stupid study. You can't make this up. What a terrible research design!