I don’t think it needs to perfect, but without seeing the data I can’t trust it. ie one of the way the results can rigged is to set lower kilometers travel for some models.

To adjust for exposure, the number of cars involved in a fatal crash were normalized by the total number of vehicle miles driven, which was estimated from iSeeCars’ data of over 8 million vehicles on the road in 2022 from model years 2018-2022.

Gived the number are estimated, how can we trust them?