Major hazard facilities, such as nuclear power stations or chemical manufacturing sites, frequently engage in cost benefit analysis (CBA) of measures designed to prevent rare but catastrophic events. These analyses are beset with difficulties, among them the difficulty of costing disastrous events and the difficulty of estimating their probability (likelihood or frequency in some models). But CBAs in these circumstances have a far more fundamental problem: they rely on a calculation of consequence times probability. This paper argues that that this figure is quite irrelevant to the owner of a hazardous facility and of absolutely no use in deciding how much to spend on disaster prevention. It further argues that owners should stop thinking of catastrophes as chance events and think of them instead as caused by failures of control systems, and therefore preventable by ensuring that controls are
working as intended.