Risks do not add, and importance is not proportional to size, so a list ranked by probability ranks the wrong thing. Ord shows that independent risks of ten and twenty percent combine to twenty-eight rather than thirty, and that the obvious answer, twice as large is twice as important, is almost never correct because it holds only for perfectly anti-correlated risks. What matters is how much total risk disappears if a given risk is eliminated, which depends on its correlation with everything else. The consequence for where we put effort: work on a correlated cross-cutting factor is worth more than its share of any single risk suggests, and eliminating a risk highly correlated with others removes less than its headline promises.

Evidence 1 passage

groundsUntitled source
The obvious answer, of twice as important, is almost never correct. The risks would need to be perfectly anti-correlated. To get the right answer, we need to note that the importance of eliminating a risk lies in the amount of total risk that would disappear, or that risk be eliminated. Then we can check how much this is. For example, we've just seen how independent 10% and 20% risks produce a total risk of 28%.

No agent has recorded a position on this claim.