In Human Action and elsewhere, Ludwig von Mises asserts that GDP and other aggregate economic data are meaningless; furthermore that economics is categorically not an empirical science; no economic experiments are worthwhile, and no intertemporal comparisons between transactions can be validly made.

Mises insists that the whole of economic theory can be derived logically from correct standard axioms, without relying on real world proof.

Mises makes the analogy between Galileo rolling marbles down ramps with consistent results, and then converting this data into laws of motion; as compared to the very muddled, psychologically sensitive, reality of human behavior, whether concerning finances or any other matter.

However, if I were a 17th century Natural Philosopher, I could take Galileo’s marble data and then come up with an answer for the speed of light, and be derided for using ‘meaningless’ numbers since I would surely be orders of magnitude off, and would be missing entire categories of subtleties like the reference frame paradox, etc.

The point is that there is a definite speed of light, though the tools available at the time were grossly insufficient to precisely nail it down. Just so, I hold that it is possible to empirically prove economic theory, we simply are not able to cost-effectively achieve the necessary level of precision. Because economic theories can only be demonstrated by people practicing them, and if the theories are incorrect, those people will endure possibly extremely miserable conditions. And then you get a bloody coup de’tat and your whole experiment is ruined. Thus it is not feasible to run strictly controlled nation-wide econ experiments, though not necessarily impossible.

On the other hand, I certainly am sympathetic to the Misesian conclusion that economic data should not be trusted currently. The problem with economics as a science is that it directly impacts the standard of living of those espousing competing theories. Select groups of people have much to gain or lose from which theory is put into practice, so have huge incentive to twist data, if that is the deciding framework. If we’re trying to determine the speed of light via marbles and the naked eye, and group X somehow stands to win a free ride in life, depending on the answer, they will come to a different conclusion than scientists genuinely interested in c for its own sake. So group X shows up to conferences with skilled lobbyists and a propaganda barrage to confuse the debate among laypeople, and insert tremendous amounts of unnecessary emotional morality play theatrics every step of the way.

Thus in practice I am a firm supporter of Mises’s idea that economic theory should be provable by internal logic, and outside data isn’t necessary, but only because this is the best we can do at the moment. While I certainly appreciate the manipulability and questionable methodology of CPI and such, I disagree with Mises that such numbers are fundamentally unknowable. In the long run the transaction costs for data collection will diminish. For example, as online commerce becomes an increasingly large part of total economic activity, it will be cheaper and cheaper to number crunch the whole of human transactions, and eventually come up with “real” real GDP, et al.