This paper was stimulated by an assertion made by Cornell, Noll, and Weingast in their excellent discussion of safety regulation (1976). Having observed that complete information about the hazards associated with all products and occupations is not and, perhaps, cannot be made available, they note that the literature on statistical decision theory shows that investment in research designed to reduce our ignorance about product and workplace hazards should lead to better decisions (assuming, of course, 'flexible decisionmaking procedures, overseen by institutions capable of responding expeditiously to past mistakes') and conclude with a plea for caution in dealing with 'unknown hazards.' (Cornell, Noll, and Weingast, 1976: 469). This is sound observation and good advice. However, they further suggest that what this means is that regulatory decision-makers ought:
... to adopt strategies that avoid the worst logically possible outcomes, thereby minimizing the maximum possible loss, no matter what the likelihood that the maximum loss will actually occur.
They also imply that this suggestion is somehow justified by political practice:
The rationality of minimax-regret is highly controversial, especially among economists; however, political scientists have produced convincing evidence that this decision rule is often consistent with observed political behavior (1976: 469-470).