Tag Archives: Knight

More on Radical Uncertainty

Gabriel Mihalache has criticized the views of myself and others on radical uncertainty as follows:

Some people wrongly interpreted Caplan’s point as being one about markets, so they jumped at a chance to criticize a set of complete, contingent markets, but a) this is not about markets, but rather about agents; and b) neoclassical economics can be done with incomplete markets or no markets at all!

Contingent claim markets are used in models of representative agents, so I am not sure where this criticism quite fits. The problem that I have with contingent claim markets and the use of representative agents in general equilibrium theory is far too expansive for a blog post. Similarly, I do not want to get bogged down with other elements of GE theory.

First, I would point out that the world is non-ergodic (to use a term of Doug North, Paul Davidson, and others). As the quote from Keynes in my previous post as well as the work of Schumpeter on creative destruction indicates that there is no probability distribution that exists for invention, innovation, etc. Similarly, as Doug North points out, economists treat uncertainty (as defined in the Knightian sense of the word) as though it is a rare case, when in fact, “it has been the underlying condition responsible for the evolving structure of human organization throughout history and pre-history” (Understanding the Process of Economic Change, Douglass C. North, p. 14).

Thus, ignoring the misuse of uncertainty in the general equilibrium framework, let’s use the classical example of risk and uncertainty from microeconomics. An actuarially fair insurance premium would be such that:

Premium = p*L

where p is the probability of the event and L is the loss. (We can expand this to include a risk premium, but it would not embolden our analysis). Of course, in reality, there are cases where both p and L are unknown. Suppose, for example, one wanted to purchase insurance against the risk of the price of a given commodity falling over an extended period of time. What is the likely price of that commodity 5 years hence? 3 years? 1 year? 3 months? What is the probability that the price will fall? As Keynes would say, “About these matter there is no scientific basis on which to form any calculable probability…”

I am in no way trying to argue that models or risk and uncertainty should be abandoned. They are clearly useful in cases in which the probabilities and potential losses are explicitly known. However, we would do well to recognize that the world is not ergodic and that always and everywhere modeling it as such is an impediment to our understanding of complex human interaction.