… is from pages 101-102 of Anthony de Jasay’s brilliant 1998 book, The State  (original emphasis):
What could be more innocuous, more unexceptional than to refrain from intervening unless the cost-benefit comparison is favourable? Yet it treats the balancing of benefits and costs, good and bad consequences, as if the logical status of such balancing were a settled matter, as if it were technically perhaps demanding but philosophically straightforward. Costs and benefits, however, stretch into the future (problems of predictability) and benefits do not normally or exclusively accrue to the same persons who bear the costs (problems of externality). Therefore, the balancing intrinsically depends both on foresight and on interpersonal comparisons. Treating it as a pragmatic question of factual analysis, one of information and measurement, is tacitly taking the prior and much larger questions as having been somehow, somewhere resolved. Only they have not been.
In short, it is philosophically naive – scientifically misguided – an exercise in scientism – to suppose that counting up dollars observed to accrue to a side that the investigator chooses to label “benefits” and then comparing that number of dollars to the number of dollars observed to accrue to a side that the investigator chooses to label “costs” is sufficient to be a sound method for objectively deciding what government policies should be enacted or not.