The car safety tests that are required for every new mass-market passenger vehicle are flawed in many ways. Here is a list of the most obvious flaws (please point out any that I’ve missed):
- There has been no research to make accurate crash-test dummies to represent women and children, and I believe that there has been no research to make crash-test dummies to accurately represent people of racial groups that are not common in the US. Basically the medical research used to make crash test dummies was performed on male cadavers that were readily available in the US.
- The standard tests involve a direct collision with a centrally targeted stationary object, a direct collision with an offset stationary object, and solid objects (representing cars) hitting the vehicle from the read and the side. These simulate crashes where there is little or no attempt made to avoid the collision, they are probably really good for protecting drunk drivers. But any sane and sober driver is probably going to make some effort to avoid the collision and the resulting impact will not be at a multiple of 90 degrees. Note that when a car directly hits the side of a moving car it is quite different to hitting the side of a stationary car (which is what is tested).
- There are no standard tests for the probability of a vehicle rolling in the event of a crash or of what would happen to the occupants if it was to roll. Rollover crashes are among the most dangerous…
- The tests do not take into account the ability of the driver to avoid a crash or minimise the damage. The ability to avoid crashes is a major advantage for cars with a low center of gravity, AWD, and traction control. It’s a major problem for vehicles with a high center of gravity and with tires that are not designed for road use (IE 4WD/SUV vehicles).
But generally the crash-test results are of some use provided that you start by looking at the results from vehicles that have good safety features such as the Audi Quattro, the AWD version of the VW Passat, a Mercedes with 4MOTION, or any other vehicle with constant four wheel drive, road tires, four wheel traction control, and a low center of gravity.
The RACV (the main car owners advocacy organisation in Victoria and also a major car insurance company)  has published the used car safety ratings report . This was produced by the Monash University Accident Research Centre and is based on the analysis of 3,000,000 crashes reported to police in Australia and New Zealand. Results are only available for cars which have been in common use on Australian and New Zealand roads for some time (so there aren’t many entries for vehicles that are less than 5 years old or for particularly expensive vehicles).
The report also includes estimates on the purchase prices of some of the safest vehicles. A vehicle that is significantly better than average can be purchased for as little as $5000!
Now if you want to buy a new vehicle then choosing the latest version of a model that has rated well on the used-car tests should be safe if the new car crash tests also report good results. It seems likely that the latest Mazda 6 or VW Passat will also rate well on the used-car tests in a few years time. It’s a pity that the report didn’t note which of the vehicles that rated well have models that have good features to avoid collisions such as EBA, ABS, AWD, and traction-control.
A friend who is active in the free software community recently had a very lucky escape from a significant crash. From his description I doubt that car safety features had much to do with him escaping without injury, I think that it was mostly luck. While his car did have a good range of safety features (and was rated well on the used-car tests), a high-speed collision that involves a truck can easily result in a car being squashed flat. I have already sent him the RACV link which he is using as part of the process to decide what new car to purchase. But I think that this information needs to be spread more widely.
I have not searched for information on such analysis of crashes being performed in other countries, please leave a comment if you know of any good research that will be useful for other people. One thing to note however is that given the global scope of car manufacturing, results from one country will have some validity in others. I expect that a VW Passat that is sold in Germany or the US will be almost identical to the Australian version.