The Socialist Calculation Problem is not about Computation or Knowledge Expressibility
With the increasing temptation produced by the growing computational capacity and the resurgence in popularity of the socialist ideas in many places, the socialist calculation problem may suddenly turn from a past curiosity into something that urgently needs Remarkably, the issue is still largely misunderstood even by the most highly-qualified people.
Yesterday, Ethereum creator Vitalik Buterin tweeted that as far as the failures of socialism are concerned, talking about “calculation” is a distraction herebecause even if its solution is very hard to compute, similar problems are solvable and solved in a satisfactory manner.
To Buterin, something else must then be the root of the problem, and he echoes F. A. Hayek invoking the expressibility of knowledge.
While Buterin does not imply that this problem with central planning could be solved, someone may wonder if something like AI could decode the tacit knowledge and finally make socialism workable. But is that all there is to the socialist calculation problem?
Socialist calculation problem: background
The first economist to have come up with the socialist calculation conundrum was Ludwig von Mises. Mises’s point about socialist calculation was that even if a socialist economy retained markets and prices for final consumer goods, it would not be able to rationally allocate inputs to production projects in the absence of private property in and prices for the resources (inputs). The reason for this is that decision-makers would lack a common profit-and-loss metric to compare possible projects.
Hayek tried to explain what Mises meant through his famous tacit local knowledge argument, however, I believe that the problem he identified is largely separate from Mises’s original meaning.
To Hayek (and Buterin who relies on him), effective central planning is impossible or at least very hard because the planner could never get access to the whole knowledge needed to compute the plan. This means both that consumers cannot adequately explain why they prefer some things over others and producers struggle to fully convey how they succeed in doing what they do. If only those obstacles could be overcome, the thinking goes, the central planners — armed, perhaps, with powerful computers — could easily do their job.
Let us embark on a thought experiment that would, hopefully, show that this reasoning is misguided.
Central planning in Langelandia
Imagine a country (let us call it Langelandia after Mises’s biggest critic Oscar Lange) where government-issued money can only be used to buy consumer goods. At the beginning of each year, Langelandia’s central planning board decide the things to be produced during the year and their quantities, as well as which resources are going to be used for which consumer goods.
At the end of the year, the board collects the information about all the quantities of consumer goods purchased by people and the prices of transactions. Suppose that after one year, the board decides to introduce a major change in the product mix of Langelandia. For instance, suppose that the board decides that electric scooters are the transport of the future.
The board plans to produce a lot of scooters the following year at the expense of some other uses of aluminium, plastic, batteries and other inputs that go into producing them. Just as a hypothetical, imagine that this results in fewer aircraft parts, electric bikes and laptops available for the Langelandian consumers.
A year later, the central planning board reconvenes to decide whether the change is worth preserving or even doubling down upon. What economic data does it have at its disposal? The answer is that it can compare the set of quantities and prices from the baseline year with the set of quantities and prices from the year with more electric scooters.
Here, we arrive at the crucial idea. Just comparing two sets of consumer good prices and quantities sold provides the central planners with no rational basis for deciding whether the new production mix is preferable.
In contrast to the central planning case, if the hypothetical boost in scooter production were attempted in the context with prices for inputs preserved, it would be a no-brainer to rationally compare the changed production mix to the preceding one. It would be just a matter of seeing whether the entrepreneurs who reallocated the inputs to scooter production managed to get higher profits than the enterprises from which they bid away those inputs.
The economy does not compute
What the Langelandia hypothetical shows is that the problem is not about gathering knowledge and computing over it. The market economy is not a substitute for a giant oimniscient computer. If it were such, it would establish all the prices simultaneously, like it happens in the unrealistic mainstream economic models.
However, in reality, the market process unfolds through incremental changes brought about by entrepreneurs who never try to reallocate all or most inputs at the same time. They just succeed or fail in their particular attempts at diverting inputs to different uses compared to the past.
What exactly entrepreneurs tinkering with particular reallocation attempts do that makes the process fundamentally different from central planning and not just different in effectiveness is a separate question which, perhaps, cannot be answered in satisfactory manner at all. However, this should not stop us from accepting that the computation metaphor is just a non-starter for the economy. And that the lack of inputs for computation is not the real problem, either.