> Those are issues of centrally planning the entire economy in 5-year cycles.
This is not the only constraint on method. They had some ways to make predictions about future demands, they had some optimization criteria.
> I don't know of what centrally planning a part of the economy would look like, nor what planning by a democratically elected government, as opposed to a self-appointed one, would be. However, as I noted, political power always tries to grow, so the moment you place that much leverage in the hands of the government, you could expect to end up with a totalitarian system, which is my primary fear of central planning.
I agree, this is a problem. But at my opinion, its not a problem of central planning (viewed as algorithm of optimization of economy), but problem of society, which have no idea how to combine in one system democracy and central planning. Nevertheless its a problem.
> It seems to come naturally to distribute the planning computation among independent agents.
Yes, but everything comes at cost. By delegating planning to independent agents you lose ability to find optimal solution of optimization problem. Such a distrubuted planning computation gives all pluses of distributed computation, its great and, probably, unavoidable, because of computational complexity. But such a system will not give us the best solution. We hope that solution will not be the worst.
Practice shows, that such a solutions is good enough, but will such a solutions be good enough in competition with supercomputer crunching numbers and searching for the best solution?
AFAIR, Kahneman told a story about lesser managers who failed to behave risky enough, because from their subjective point of view such a risks was too high. But from point of view of higher level manager those risks were acceptable.
I'm not saying that existing methods is bad, I'm wondering if there are better methods to plan, and is there are possibility to see them implemented in reality somewhere in the future.
This is not the only constraint on method. They had some ways to make predictions about future demands, they had some optimization criteria.
> I don't know of what centrally planning a part of the economy would look like, nor what planning by a democratically elected government, as opposed to a self-appointed one, would be. However, as I noted, political power always tries to grow, so the moment you place that much leverage in the hands of the government, you could expect to end up with a totalitarian system, which is my primary fear of central planning.
I agree, this is a problem. But at my opinion, its not a problem of central planning (viewed as algorithm of optimization of economy), but problem of society, which have no idea how to combine in one system democracy and central planning. Nevertheless its a problem.
> It seems to come naturally to distribute the planning computation among independent agents.
Yes, but everything comes at cost. By delegating planning to independent agents you lose ability to find optimal solution of optimization problem. Such a distrubuted planning computation gives all pluses of distributed computation, its great and, probably, unavoidable, because of computational complexity. But such a system will not give us the best solution. We hope that solution will not be the worst. Practice shows, that such a solutions is good enough, but will such a solutions be good enough in competition with supercomputer crunching numbers and searching for the best solution?
AFAIR, Kahneman told a story about lesser managers who failed to behave risky enough, because from their subjective point of view such a risks was too high. But from point of view of higher level manager those risks were acceptable.
I'm not saying that existing methods is bad, I'm wondering if there are better methods to plan, and is there are possibility to see them implemented in reality somewhere in the future.