Perrow's thesis was that systems needed three characteristics to generate normal accidents: to be very complex (beyond any single human's detailed comprehension), to be tightly coupled (meaning that the system can change its behaviour quickly) and to have catastrophic potential.
Pressurised water designs are inherently complex and coupled because the necessity to circulate fluid at all times without failure, making them complex (lots of moving parts and failure modes) and coupled (failures can rapidly propagate their effects). They also produce plentiful radioactive byproducts.
A counterargument is that non-PWR designs in Gen-III+ and Gen-IV actually address some or all of these characteristics. Many of them are much simpler, reducing complexity. Many of them are design to operate passively, reducing complexity and coupling. And some of them reduce radioactive byproducts.
That's a pretty interesting metaphor, because while the newer car is indeed much safer, there are many types of accidents where the driver will be killed no matter what car they're driving. Perhaps they'll still have all the body parts attached in the Honda though ;)
One of my favorite books is "Normal accidents" by Perrow, where atomic plants are the perfect example of high-risk systems which invite system accidents.
Their dedicated chapter is comically horrific. e.g.:
"In 1978 a worker changing a light bulb in a control panel at the Rancho Secco 1 reactor in Clay Station, California, dropped the bulb. It created a short circuit in some sensors and controls. Fortunately, the reactor scram controls were not among those affected, and the reactor automatically scrammed. But the loss of some sensors meant the operators could not determine the condition of the plant, and there was a rapid cooling of the core. [...] did not in this case damage the core. But this is probably only because the plant had been operating at full power for less than three years. A spokesman for the NRC said: “If it had been 10 to 15 full power years, instead of two to three, which it was, that vessel might have cracked.” A cracked vessel would result in a loss of coolant and a meltdown; no emergency system would be available to cool the core."