But then I can make the same claim again: why should we be confident we’ve got the percentage of the capacity right?
I think even if we’re not confident, bounds on welfare capacity can still be useful. For example, if I know that A produces X net units of good (in expectation), and B produces between Y and Z net units of good, then under risk-neutral expected value maximization, X < Y would tell me that B’s better, and X > Z would tell me that A’s better. The problem is where Y < X < Z. And we can build a distribution over the percentage of capacity or do a sensitivity analysis, something similar to this, say.
I think even if we’re not confident, bounds on welfare capacity can still be useful. For example, if I know that A produces X net units of good (in expectation), and B produces between Y and Z net units of good, then under risk-neutral expected value maximization, X < Y would tell me that B’s better, and X > Z would tell me that A’s better. The problem is where Y < X < Z. And we can build a distribution over the percentage of capacity or do a sensitivity analysis, something similar to this, say.