Is it possible we are wrong?
What we see in any dramatic test with a dummy in the car is that these safety improvements protect the dummy. Is that a valid test of the real world.
I have one piece of data that suggests the real world is not like the one we see with crash test dummies.
Japan has had a relatively stable automobile fatality rate for the past few decades. In roughly 1993 Japan went from a country where seatbelts were only required on highways as of December 1992 to a country where seatbelts were required at all times in all front seats in January 1993. The Japanese did as they were told. 100%.
This is what leads me to make another suggestion. Here is another chart that shows auto fatality by age. It is clear that people under 35 have a much higher auto fatality rate while driving. Under 20, the situation is even worse.
Hypothesis: while safety devices definitely work, the reality is that making driving more difficult for young drivers is far more important in reducing auto fatalities.
In Japan, a drivers license requires many hours of training and at least three hours of driving in the drivers test.
The greatest drop in U.S. automobile fatalities occurred in the past three years. A drop of nearly 25%. This is due entirely to severe state restrictions on teenage drivers.
It may be that drivers' habits are more related to safety issues than auto technology.