Risk studies exist because humans are infamously not capable of assessing risk without objective data.
Which is why Fletch referred to a study to support his decision to not worry about it. We decide, then we rationalize, then we study. That's what I do, and studies show we all do.
Of course it is pretty safe to say there are many much greater hazards being faced than having our pretty face marred by an air bag.
On that point, the actual number of air failures are, of course, higher than those noted. The numbers used by the NHTSA are only those proven, that can be pointed to with little controversy. Cases in which the air bag failed but damage from the crash was so great it wasn't investigated cannot be included. With a national autopsy rate of 8.5 percent it is unlikely even half of those dying in car crashes are closely examined to determine the sources of each injury.
Not that the real number would change the conclusion that there are greater risks out there. I'm just noting how flawed NHTSA data can be. They try to work around it but it is difficult when the raw data is missing or flawed.
I've seen this first hand in how police reported seat belt use back in the day. In one area I got to look at many of the crash reports and even talked to some cops about crashes. My interest had more to do with the damage to the vehicle so I was looking at the damaged vehicles in the compound when I noticed that some seat belts were fastened, including one that I knew to be a fatal (small town). Most were either cut or unfastened so I asked about that one I saw fastened, did someone fasten it after the crash? Nope, no one had messed with the seat belt that is how it was found. The report concluded that the lone driver was not wearing their seat belt, that they had fastened it and were sitting on it which is why they were ejected during the rollover. To my untrained eyes it looked used, and it was loose so I wasn't so sure.
So I looked at other vehicles in the compound (they had years worth) and those reports I could get.
We know that three point seat belts do not work in rollovers. They are not designed for roll overs, and you will be ejected even if wearing such a seat belt (search youtube for video of three point failures). But in the reports I saw almost every ejection was recorded as not wearing seat belt, even when the seat belt was found to be fastened.
There was also a problem in recording the seat belt buckle or latch failure. That area had none, even though there was a person who told the police they had their belt on before the crash.
In my discussion with a cop about these reports he said that you wouldn't be ejected if you are wearing a seat belt properly and that an examination of the belt will show heat damage where the belt is locked into place if the person was wearing the belt.
He had accident investigation training but thought that seat belts worked in roll overs, and was unaware of seat belt failure modes, such as the buckle opening, that would not result in any obvious signs. Also he had many other duties so really could not stand around pondering possible failure modes that might be beyond his training.
His reports, like all such reports, were then used to tell us about seat belts, their effectiveness and that seat belts rarely failed if they were used.
Today accident investigation training is much more advanced and no one expects a three point to work in a rollover but they thought they had great training back in the day as well. We can only train to look for stuff we know about and agree should be looked for. The numbers we collect will always be flawed. Consider the study that looked at "speeding" as a cause of motorcycle crashes and found that only a very tiny percent had actually being traveling faster than conditions would allow. Our raw data is often flawed which is why people can make careers out of studies.
And which proves once again that it isn't good to see how the sausage is made. All you need to see is one or two reports being submitted with errors in them to have you question the studies based on them.