• 420stalin69
    ·
    edit-2
    1 year ago

    The “calculation problem” is absurd because it’s equally true for a market system anyway since any Turing system is only capable of what any other Turing system is also capable of.

    The idea that a central planner lacks complete information about an economy is true, but it’s equally true for any market. A market cannot have complete information about itself even in aggregate across the entire market. Provably true via the Turing argument but also empirically obvious since prices in any and every true market are constantly fluctuating even in the absence of events that should affect prices and never reach anything that even resembles an equilibrium.

    Further, it’s empirically proven that higher “allocative efficiency” is achieved under administered rather than dynamic systems.

    The first flaw in the “calculation problem” is that it implicitly assumes that markets achieve economic equilibrium… which is absolute horse shit they obviously do not achieve equilibrium prices. So it’s wrong in the sense it’s equally true for markets.

    The second flaw is that empirically it’s demonstrably false since it’s been shown empirically that planned systems do in fact achieve higher allocative efficiency. So it’s wrong in the sense it is contradicted by actual observation.

    The third flaw is that it makes the enormous conflation of assuming that because perfect efficiency is impossible that a system is therefore impossible. E.g. because you can’t swim at an Olympic level you’re going to drown in a pool. Actually perfect efficiency isn’t required for a system to function, obvious because perfect efficiency has never been achieved and yet we are still able to eat food.

    The nail in the coffin of the “calculation problem” is that it believes that only markets can approach efficiency since only the market in aggregate can have perfect knowledge about an economy so a price mechanism approaches efficiency… but that’s exactly what linear programming does too - it constantly moves towards while never reaching perfecting efficiency in polynomial time - and in fact linear programming provably (in both the mathematical and empirical senses) approaches efficiency far faster than a “drunk-walk” market mechanism.

    You don’t need super advanced computing to do this. GOSPLAN did it using pen and paper. Linear programming isn’t actually difficult mathematics.

    All of the above points were made about one hundred years ago and the “calculation problem” has never been taken seriously as even existing as a problem outside of the Mises institute, which is exactly why Reddit dweeb fascists love to cite it.