Taken from
an online discussion of Austrian economics, in which someone is attempting to prove that the law of diminishing marginal returns can be known
a priori.
The law of marginal utility is indeed universally true. Any person will value less of a good more than if he had large quantities of it. For instance, the cigarette addict will value his last two cigarettes higher than if he had all twenty in his pack. He would also value two cigarettes far less if he had a whole carton of them at his disposal. Of course he would prefer to have more of them, which is your point and actually a consequence of the law. The law of marginal utility states that with an increase in quantity, the value of each individual item decreases. Thus, he can put each one to its next best use:
A.) Smoking one now.
B.) Smoking one in an hour.
C.) Magic tricks with a cigarette.
D.) Having a cigarette for a friend.
E.) Smoking one in two hours.
...
Or, if he's down to his last two, all other uses will be put aside so that he will prefer to smoke one now and the other in an hour, instead of using one for a magic trick, saving one for a friend, etc. The reason he would want more is so that he could achieve all of his desired ends.
Counterexample: suppose your friend will be visiting in an hour and you would like to smoke with him. If you have two cigarettes, you hold onto them so that in an hour each of you will smoke one. If you have only one cigarette, you prefer to smoke it immediately.
Thus, we have a case where your preferred use for your first cigarette is different across the cases where you have differing numbers of cigarettes. Hence, a possible counterexample to the law of diminishing marginal returns, meaning that it cannot be known
a priori.
I'm not familiar enough with Austrian economics to know if this is the argument that Mises would use to "prove" diminishing marginal returns, but in any case this argument is bunkum.
(Neoclassical economists have a far better way of establishing the law of diminishing marginal returns: we simply assume it. But at least we're honest about the fact that we're assuming it and have the tools to relax that assumption if necessary. Fortunately it seems to hold up pretty well in the real world).