Promulgated by John Adams & Gerald Wilde, as explained in their books "RISK" and "TARGET RISK"
Adams and Wilde explain the failure of seat belts to save lives with the Risk Compensation Theory (RCT). RCT is based on the hypothesis that "protecting car occupants from the consequences of bad driving encourages bad driving." to such an extent that the beneficial effect of the safety device is completely negated. (Adams 1982, abstract)
Their hypothesis is supported by persuasive evidence. In the Munich Taxi Cab experiment, (Reported by Wilde) drivers with ABS (anti-lock brake system) were found to compensate for the extra safety of ABS by driving more recklessly than drivers without ABS. Wilde (section 5.5) reports evidence that drivers are more careful when bad weather increases driving hazard.
Wilde suggests that driving care is a function of "The perceived level of risk". Chapter 4.2) So collapsible steering columns and crumpling bodywork would probably not affect driver safety behavior. Snug seats, better roads & better brakes probably would.
Wilde states that seat belts and alcohol are perceived as affecting
safety and are consequently subject to the hypothesis.