A hedge fund manager who is short Tesla just published a piece on Medium arguing that Teslas are four times as dangerous as luxury cars in general, contrary to Elon Musk’s statement that they’re four times as safe. He used data from the Insurance Institute for Highway Safety, which rates these things in terms of driver fatalities per million vehicle-years, and says that the average is 13 per million vehicle-years, while Tesla has already had 11 in 265,000 vehicle-years, which is a rate of a little over 44 per million vehicle-years.
This analysis is rife with problems.
First off, the author says,
“Along the way, we discovered several errors in the database, likely due to fact that Tesla is not assigned a unique model code but is rather part of an obscure basket of “Other domestic manufacturers” with Studebakers, Hudsons, Packards, and others. Worse yet, that isn’t even the only “other vehicle” basket. We found Tesla accidents coded correctly, Tesla accidents coded incorrectly, and Tesla accidents seemingly completely missing.”
Given that unreliability, is it not as likely that there are accidents coded as Teslas which actually were not? A serious scientist and thinker would discuss each possibility for error, but he does not mention this one.
Second, the author talks about how reliable the statistical sample is, because they have almost 3 million vehicle-years as a sample size.
The problem is, it’s not the vehicle-years that are the sample size, it’s the number of fatal accidents, because fatal accidents can occur for many reasons, some unrelated to car safety. 11 (or 13) is simply so small a number that any number of factors causing fatalities could skew the results wildly. Over a large sample size, that sort of thing will even out, which is of course why scientists are more confident about large samples– because over large numbers, skewing factors cancel each other out. But as noted, that’s just not the case here. What if these particular years happened to see 11 accidents with Teslas that would have been fatal regardless of the car’s maker, but not so much with the others? What if Tesla drivers, unlike luxury cars in general, skewed particularly young and male (a demographic more prone to aggressive driving and accidents) or these 11 happened not to be wearing seat belts? Even more fatal to this finding of significance, what if the fatal accidents clustered around the early part of the time period, such that any deaths as the result of the car itself were the result of problems that Tesla has now solved?
It’s particularly strange that a professional investor would make this mistake. With so small a number, it’s like seeing a week’s worth of market swings in a stock’s price and thinking that they must be reflective of changes in the true value of the stock.
Mind you, I have no opinion on the actual safety of Teslas, nor on the company’s honesty or financial health (though it sounds like Musk is on thin ice either way with his puffery about Teslas being much safer than the average luxury car), nor even on the author’s short of the company’s stock (though a successful short requires a company’s stock to go down dramatically in the relatively near future, and given the promotability of Tesla’s story and the tiny number of 11 driver fatalities per 265,000 vehicle years, that’s not sufficiently certain for it to be a bet I’d be willing to make).
I do believe, however, that the author has tried to spin gold out of insufficient straw. To me, this evidence is simply not yet strong enough to support his conclusion. Perhaps other evidence exists out there which would firm it up. Perhaps not.