Safety is complex. Are some cars safer in crashes than others? Yes. Is it a lot? Depends on the crash. Are some cars safer in all crashes than other cars? No. Are crashes the most significant factor about safety? No.
A recent SAE study showed that the #1 factor is matching the vehicle to the driver. Older drivers in larger vehicles tend to have a higher accident rate - as we age, our ability to tangibly grasp the size of a vehicle declines, and the larger the vehicle, the less accurate we are in lane positioning, etc.
Once we retire, we have the luxury of picking when we drive so that we're on the road when there's less traffic. Way more important than how many stars of safety rating a car has.
About 15 years ago, I wanted to set my mind at ease. I don't like driving big vehicles: I don't feel I can see out of them well, and it bothers me if a vehicle doesn't respond quickly to my driving inputs. I have never had normal depth-perception, and while it's rare, it happens occasionally that I misjudge distances and must respond to avoid problems caused by my own error. Quick handling and responsive acceleration are really important in that situation. I don't need a hot rod, but the slowest automatic-equipped cars with soft suspensions are simply inadequate for me. Other people will have different needs, of course.
Anyway, so my tendency is to drive rather small cars. But, was I taking my life into my hands?
I pored over the NHTSA safety report, seeking the statistics of likelihood of survival in this or that crash.
Worst case: The front-seat driver of a one-star car was 46% more likely to die in a 55 mph front-to-front full-contact collision than the driver of a five-star car. If the five-star car were huge and the one-star car small, add a bit less than 10% to that. Whoa...sounds scary. But, I read onward.
If you compared a 4-star and 4-star, one big and one small, the NHTSA claimed the difference in survivability rates for the same collision were a few percent "either way". In other words, there are intricate factors involved such as whether the bumpers are at the same height, etc, which render the "star" system to be fairly gross.
What about other kinds of accidents? The NHTSA claimed that the reason they focused on the head-to-head collisions is that "they represent a worse-case scenario which results in the highest probability of death" and are roughly equivalent to a single vehicle hitting a concrete wall or light post head on. In all other accident types (glancing blow, side-hit), speeds statistically tended to be lower and fatality rates lower and "relative differences between vehicles are both less significant, and more difficult to ascertain, therefore we do not allocate much budget to them". Also "head on collisions above 55 mph become substantially equal in terms of likelihood of death irrespective of vehicle, and head on collisions below 40mph so infrequently lead to death that we deem all vehicles equal".
I haven't read more recent reports in as much detail, but I've skimmed them and seen similar words.
Everybody has to make their own choice, but I've decided that the accident-avoidance aspects of small nimble cars, for me, outweigh the not-always-correct safety benefits of "large", and for me, as long as I avoid the bottom two star ratings, I figure my risks in a crash probably fall off my top 5 risks of driving list.
BTW, early airbags were kind of scary - they could cause permanent injury in accidents that otherwise would have caused none...an old pal of mine became a paraplegic when his car's airbag went off in a 25 mph collision. However, since about 2005, by law, airbags deploy with different levels of force depending on both the speeds involved in the collision and the weight of the passengers in the adjacent seats. Good thing, in my view that makes them ultra safe. Except for people not belted in, kids or pets on laps, etc.