One's a mirror of the other. Stocks and flows are primal values; prices are dual values. And if you live in the nice kind of economy where there're no fixed co

One's a mirror of the other. Stocks and flows are primal values; prices are dual values. And if you live in the nice kind of economy where there're no fixed costs or nonlinearities, you can translate between the two without any loss of whatever it is you're trying to optimize, courtesy of the strong duality. Reality is not that nice, of course, but a linear programming economy would be vastly better than what the USSR had at the time. Not that they could've switched even if they had supercomputers: too much politics invested in the material balances system. By the way: Shalizi paints probably a bit too bleak a view on the use of large-scale LP. There exist linear programming algorithms that are O(nnz(A)), i.e. scale as the number of nonzeroes in the (presumably sparse) matrix. E.g. https://papers.nips.cc/paper/2015/file/0966289037ad9846c5e994be2a91bafa-Paper.pdf https://papers.nips.cc/paper/2015/file/0966289037ad9846c5e99...