Comment by eesmith | Hacker News Reader

Comment by eesmith | Hacker News Reader

Parent Full thread

"too much optimization at the expense of robustness"

That is not unique to central planned economies. Consider how the 2011 Thailand floods lead to a global shortage of hard disk drives, or the saline solution shortage in the US when the plants in Puerto Rico were damaged by Hurricane Maria.

I think the author expressed it clearly that the planning algorithms depend on linear programming, and that non-linear effects - and I believe 'robustness' counts as such - are beyond what can even hope to achieve. The author also points out:

> If it’s any consolation, allowing non-convexity messes up the markets-are-always-optimal theorems of neo-classical/bourgeois economics, too. (This illustrates Stiglitz’s contention that if the neo-classicals were right about how capitalism works, Kantorovich-style socialism would have been perfectly viable.)

Hence why capitalist systems may also have robustness problems.

"You can have an efficient system but it's robustness (ability to withstand a failure or adapt to change) will be very low."

I think that's an observation in how centralized systems are implemented. I don't see why robustness can't, at least in theory, be included as part of the optimization process.

Beyond the factors highlighted in the essay, of course, like the inability to decide how to weight those factors, and the non-convexity.

"That is not unique to central planned economies."

Yes, however, when it happens in market economies, it is not caused by planning for greater efficiency (saving money). Nobody really decided there to be only one factory that produces hard drives; it is a consequence of hard drives being particularly difficult to produce.

"I think the author expressed it clearly that the planning algorithms depend on linear programming, and that non-linear effects"

I disagree with that. I think planning itself (algorithms) is not the issue here, I think the objective is.

"I don't see why robustness can't, at least in theory, be included as part of the optimization process."

In theory, perhaps. But in practice, it's difficult to prove to the people in charge (whoever that is) that you made the system more robust, compared to proving that you made the system more efficient. So there will always be an inherent organizational bias towards efficiency.

Airlines are an interesting example of industry which strives for efficiency, yet has strong robustness guarantees through regulation enforced by central agency. (Although in this case, it is not economic robustness.)

Maybe the robustness of the economy as a whole is given by how much choice you have in being conservative or take an experiment. So there really cannot be "robustness" as an objective, since it depends on continual experimentation with different models of operation, where most of them fail.

>Nobody really decided there to be only one factory that produces hard drives; it is a consequence of hard drives being particularly difficult to produce.

Are they? To the point that some factories in Thailand had all/most the knowledge and equipment to do so?

I'd say it's exactly they were located in Thailand for exactly the same reason a planned economy would focus on: "greater efficiency (saving money)": it was convenient and cheaper to get them made there (except when it suddenly weren't).

What I am trying to say is that it was a convergent decision to buy from a single manufacturer (although perhaps cost-driven), rather than a decision to, in order to reduce costs, have only one manufacturer.

Making HDDs is quite difficult and there is only a handful companies that can do it. If the barrier to entry was lower, there would be more different manufacturers and probability of relying on single manufacturer would be reduced. Like in most other industries.

>What I am trying to say is that it was a convergent decision to buy from a single manufacturer (although perhaps cost-driven), rather than a decision to, in order to reduce costs, have only one manufacturer.

Isn't the end-result the same?

"however, when it happens in market economies, it is not caused by planning for greater efficiency (saving money)."

I think the author addressed that, writing:

> "For the capitalist or even market-socialist firm, there is in principle a simple objective function: profit, measured in dollars, or whatever else the local unit of account is."

"I disagree with that .. I think the objective is"

I'm not sure what you disagree with. My incomplete summary? The author also goes into detail about the problems in determining the objectives. I mentioned them in passing in my last line.

"In theory, perhaps"

Well, yes. That's because it's a response to the theoretical statement you wrote: "You can have an efficient system but it's robustness (ability to withstand a failure or adapt to change) will be very low."

Do you have any evidence that this is a universal statement?

Anyone worried about possible environmental catastrophes as the result of greenhouse gas omissions must certainly be wondering what can be done to include that robustness factor in a market economy.

The impact from greenhouse gases is already either a factor in some power markets (which heavily use state of the art optimization and are among the hardest problems to solve when you look at the size versus the time it has to solve in) or planned to be included. It is indirectly there in all power markets as coal is currently becoming less and less economic.

"I'm not sure what you disagree with."

I disagree with the usual notion that the problem of planning is somehow (algorithmically) difficult, once you know the objective function.

Instead, I argue there is a tradeoff in the objective function, which is hard to estimate in theory without experimentation. This experimentation goes against the goal of efficiency (you need redundancy).

So even if you knew all the demands of all the consumers, and could calculate what to produce (which e.g. von Mises argues is impossible), I argue you would still have a problem with central planning, because the resulting optimal plan would not be robust enough to execute. I also don't think the former is a big issue with computers, to be honest. There are examples of top-down logistical planning operations in the world which are comparable to an economy of a small country.

"Do you have any evidence that this is a universal statement?"

Well, in optimization problems, usually as you get closer to the possible optimum, the number of acceptable solutions decreases. So the closer the solution to the optimum, the less leeway there is to adapt the solution to some unforeseen circumstances.

Addendum: I think many economists try to paint free market as being efficient, and that's why they argue the optimization problem is impossible to solve. However, I don't think free markets actually are efficient, but rather, make a different trade-off, which works better in practice (it also has downsides).

Interesting to hear you say this.

Power energy markets are run using LP & MILP algorithms and a huge topic right now is accounting for non-convexivities.

Yes, imagine the chaos if Walmart failed tomorrow.

Reply on news.ycombinator.com