Expensive Cheap Processing Power
Imagine that you run a factory. In the status quo, workers perform the tasks that are assigned to them by managers, who must perform a series of calculations to organize the work. The calculations are burdensome, and you doubt that the factory's organization is "optimal" (except in the sense that it is the best you can do given the limited managerial skill and processing power at your disposal).
Now an entrepreneur offers to sell you a computer to replace the managers. The computer does the same thing that managers do: it gives instructions to workers in order to organize production at the factory. But unlike the managers, the computer does not have to be paid a salary, and it may also give better instructions overall. As a result, there is a potential for improving the profitability of the factory (either producing the same output at a reduced cost, or producing more output at the same cost).
So this machine at least has the potential to be a worthwhile purchase. But of course you aren't just going to cut a check on the spot: you want to make sure the machine is actually worth the purchase price. So you ask: how do I know that the instructions generated by the machine will be the "right" instructions? After all, it would be easy enough to build a machine that gives clear, unambiguous instructions but doesn't improve the factory's performance. It could just tell every worker at every moment, "Take it easy, you're on break." Then you would have saved a lot on managerial salaries, but the factory's output would plummet. You would also get a bad result if the machine gave instructions at random - and in that case, the results might be dangerous for the workers.
The seller of the machine assures you that its instructions are much better than random. But he admits that the machine sometimes instructs workers to do things that are counterproductive. So actually, the suggested strategy is to retain enough managers to override the machine when it gives a bad instruction. You ask the vendor for an assurance that you will still come out ahead with this approach. But the vendor is honest, and he tells you that he doesn't know. It just depends on how often the machine gives a bad instruction, how many managers it takes to catch these mistakes, how bad the mistakes are, etc. It's a complicated question that can't be resolved by pointing to the machine's incredible efficiency. Terrible instructions, however cheaply they are generated, are no bargain.
And this is where we find ourselves with the economic calculation problem. We know that capitalism transmits "instructions" somewhat cheaply (although capitalism does seem to have a voracious appetite for smart, educated people to run the financial system). But this, by itself, is not a reason to prefer capitalism to socialism. You also have to assess the quality of the instructions (in this case, price signals), and the cost of monitoring them and overriding them when they go wrong. Recall that in almost every year of the Irish famine, the price system induced Ireland to be a net exporter of food. The "instruction" from the machine was: remove food from a starving nation and deliver it to a well-fed one. And no one overrode the instruction. The Irish were left to starve. A socialist is free to argue, I think, that capitalism's cheap processing power comes at too high a price.
This is another way to view the Coasian point about the size of the firm. At low levels of complexity, we may prefer to keep the human managers and dispense with the machine, because its huge processing power doesn't make up for its glaring shortcomings (in terms of bad instructions or other pathologies). So we don't run families, churches, or corporations on a market basis. Arguably the machine's benefits start to outweigh its costs at large enough scale, but this is not a question that can be answered in the abstract. It depends on how good our non-market institutions are, how much it costs to maintain them, and what value we place on, for example, Irish lives.
So the economic calculation problem is not the end of the discussion, it is the beginning. Unfortunately, that's the opposite of the way it is usually presented in economics classes, in my experience.
And even if you are convinced that capitalism is a virtual necessity in a modern economy, it doesn't follow that its dictates are irreproachable and beyond human review. There is still a debate to be had about how and when to override the allocation chosen by the price system. And a lot is at stake. That's why it's obnoxious that, again, economics instruction essentially truncates the discussion by recognizing only very limited kinds of market failures. All the time economists will say things like, "There's no externality here and therefore no justification for government action." As if economists have somehow discovered a trick that allows them to bypass human values in formulating policy. As if Irish lives don't matter.
Now an entrepreneur offers to sell you a computer to replace the managers. The computer does the same thing that managers do: it gives instructions to workers in order to organize production at the factory. But unlike the managers, the computer does not have to be paid a salary, and it may also give better instructions overall. As a result, there is a potential for improving the profitability of the factory (either producing the same output at a reduced cost, or producing more output at the same cost).
So this machine at least has the potential to be a worthwhile purchase. But of course you aren't just going to cut a check on the spot: you want to make sure the machine is actually worth the purchase price. So you ask: how do I know that the instructions generated by the machine will be the "right" instructions? After all, it would be easy enough to build a machine that gives clear, unambiguous instructions but doesn't improve the factory's performance. It could just tell every worker at every moment, "Take it easy, you're on break." Then you would have saved a lot on managerial salaries, but the factory's output would plummet. You would also get a bad result if the machine gave instructions at random - and in that case, the results might be dangerous for the workers.
The seller of the machine assures you that its instructions are much better than random. But he admits that the machine sometimes instructs workers to do things that are counterproductive. So actually, the suggested strategy is to retain enough managers to override the machine when it gives a bad instruction. You ask the vendor for an assurance that you will still come out ahead with this approach. But the vendor is honest, and he tells you that he doesn't know. It just depends on how often the machine gives a bad instruction, how many managers it takes to catch these mistakes, how bad the mistakes are, etc. It's a complicated question that can't be resolved by pointing to the machine's incredible efficiency. Terrible instructions, however cheaply they are generated, are no bargain.
And this is where we find ourselves with the economic calculation problem. We know that capitalism transmits "instructions" somewhat cheaply (although capitalism does seem to have a voracious appetite for smart, educated people to run the financial system). But this, by itself, is not a reason to prefer capitalism to socialism. You also have to assess the quality of the instructions (in this case, price signals), and the cost of monitoring them and overriding them when they go wrong. Recall that in almost every year of the Irish famine, the price system induced Ireland to be a net exporter of food. The "instruction" from the machine was: remove food from a starving nation and deliver it to a well-fed one. And no one overrode the instruction. The Irish were left to starve. A socialist is free to argue, I think, that capitalism's cheap processing power comes at too high a price.
This is another way to view the Coasian point about the size of the firm. At low levels of complexity, we may prefer to keep the human managers and dispense with the machine, because its huge processing power doesn't make up for its glaring shortcomings (in terms of bad instructions or other pathologies). So we don't run families, churches, or corporations on a market basis. Arguably the machine's benefits start to outweigh its costs at large enough scale, but this is not a question that can be answered in the abstract. It depends on how good our non-market institutions are, how much it costs to maintain them, and what value we place on, for example, Irish lives.
So the economic calculation problem is not the end of the discussion, it is the beginning. Unfortunately, that's the opposite of the way it is usually presented in economics classes, in my experience.
And even if you are convinced that capitalism is a virtual necessity in a modern economy, it doesn't follow that its dictates are irreproachable and beyond human review. There is still a debate to be had about how and when to override the allocation chosen by the price system. And a lot is at stake. That's why it's obnoxious that, again, economics instruction essentially truncates the discussion by recognizing only very limited kinds of market failures. All the time economists will say things like, "There's no externality here and therefore no justification for government action." As if economists have somehow discovered a trick that allows them to bypass human values in formulating policy. As if Irish lives don't matter.
posted by James | 8:05 AM
2 Comments:
Cosma Shalizi's rather long post on this is worth reading http://crookedtimber.org/2012/05/30/in-soviet-union-optimization-problem-solves-you/
I agree that the default presumption is slightly annoying. I guess I think markets add value if at least three conditions are met -- 1. the actual needs involved are fluctuating and hard to gauge, but people know what they want, 2. most participants are demand-limited rather than purchasing-power limited, 3. fixed-cost and other temporal issues are not too severe, i.e., the optimal choice isn't something like trains where the benefits only accrue after the original investors are dead. There are all sorts of situations where these fail, e.g., I think 1. fails for health insurance, as well as for other obvious things like public restrooms and water fountains, and in such cases there's not really a strong potential upside to market allocation... ("There are easy instances of the calculation problem")
Its also worth noting that if Irish peasants had owned the land the market allocation would have led to a different result. BY the 1840s only 5% of lands were Irish owned. If a market socialist government had passed land reform 15 years earlier the market would probably have led to a fine and morally acceptable outcome to the famine.
Market outcomes are not path independent of initial and prior conditions. Coase theorem aside friction, transaction costs and government bestowed advantages are all real things.
That said if a historically equitable property distribution had still led to people dying in droves you would still need to screw the market and redistribute food.
Post a Comment
<< Home