While trying to understand the sources of economic profit, the economist Frank Knight introduced a powerful distinction between risk and uncertainty. Knight’s distinction is more nuanced and interesting than the caricature view that his critics dismiss. In my view, the difference between risk and uncertainty hinges on the range of possible actions an investor, manager, or entrepreneur can take to achieve their desired result.

Clear choices and risk

People encounter risk when they face what I call a “clear choice,”  where an agent can specify, in advance the possible outcomes relevant to their actions. A gambler, for instance, knows that there are precisely 38 possible pockets where the roulette ball could land, and where the ball lands will influence whether his bet pays off or not. Other examples of clear choices include a fund manager buying Microsoft stock or an insurance agency insuring a building. In many simple choices, the possible states of nature are bimodal–the Patriots will beat the spread, or they will not. At other times, outcomes extend along a single dimension, such as the range of possible returns to an in Microsoft shares.

The challenge in simple choices lies not in specifying outcomes, which is straightforward, but assigning ex ante probabilities to their occurrence. Often, simple choices are similar enough that past events can be treated as comparable observations, aggregated and analyzed statistically. Investors can, for example, collect and crunch data on historical stock performance across many companies in the same industry, and use their analyses to estimate a probability distribution for the stock under consideration. Analysis of comparable cases does not provide an infallible basis for predicting the likelihood of future outcomes, but it does some basis.

Messy situations and uncertainty

Uncertainty arises out of “messy situations,” where the variety of actions an agent could take to achieve her goals is nearly infinite. Steve Balmer, to give a concrete example, could share the same objective as a fund manager–to increase the value of Microsoft’s equity. He and his colleagues, however face a daunting range of levers they could pull to improve performance: They could acquire competitors, lobby for anti-trust relief, bet big on new technologies, run experiments with new products, focus on cutting costs, or change prices to increase profits or market share, to mention just a few possible. Microsoft executives grapple with a messy situation, where they must decide which levers to pull, how hard, in which combination, and sequence. The investor in contrast faces the clear choice of investing or not.

The broader range of possible actions complicates the data required to inform action. In messy situations, managers cannot specify in advance all the possible states of nature that might influence their best course of action. The optimal combination, sequence and timing of future actions will depend critically on the interaction of many exogenous variables (e.g., exchange rates, GDP growth, input prices, exchange and interest rates, geo-political events, natural disasters) and endogenous events including technological innovation, regulation, competitive and interaction. The possible outcomes will depend on the confluence of variables at a specific point of time, which render the situation unique for agent considering what to do.

Possible actions, not probability assessments, distinguish risk from uncertainty

The difference between uncertainty and risk is not the result of differing objectives–a manager and investor could both want stock appreciation. Nor does the distinction arise from the level of external turbulence–Microsoft faces the same level of unpredictability whether one is running the company or investing in it.  Nor, in any fundamental sense, does the difference hinge on our ability to assign probabilities to events: Agents facing messy situations cannot assign probabilities to possible outcomes, but that is a mere byproduct of the more fundamental challenge–they cannot even specify the relevant outcomes in the first place.

The relevance of outcomes is defined by the range of actions an agent can take to achieve her objectives. In clear choices, the range of actions is limited and well-defined–a fan bets or not–but mind-boggling in messy situations–the coach juggles countless interactions of trades, training, strategy and tactics to win. The following thought exercise illustrates the distinction between risk and uncertainty.

Imagine Warren Buffett goes to the mountain and God hands him a tablet with the correct probability distribution of possible market caps for Microsoft a decade hence. Assuming Buffett’s investment horizon is ten years, this information is most useful in deciding whether or not to invest. For Steve Balmer, this information is practically useless. He could quit of course, or buy more options based on the divinely revealed probability distribution. But if he decides to stay, he would need much more granular data on how competition, technology, regulation, macroeconomics and other factors might interact to create specific opportunities or threats. Only this more detailed information would allow him to take precisely the right actions in the right combination at the right time.

Why it matters

Knight’s distinction is more than an academic quibbling over definitions. First, Knight reminds us that we cannot “tame” uncertainty through reductionist definitions and statistical models. There are, of course, clear choices such as insurance policies where statistical analysis of past cases can produce useful insight.  Many of the most important situations faced by regulators, politicians, entrepreneurs, and managers are, however, unique. Historical antecedents may raise good questions or suggest pitfalls to avoid, but statistical analysis is useless.

Second, Knight’s point of view raises the important question of how we can act despite irreducible uncertainty. Knight emphasized judgement, a slippery construct best summarized as an agent’s ability to frame a messy situation in a way that permits action. Judgement is one way to manage uncertainty, but not the only or best. I will discuss others in subsequent posts.

Finally, Knight’s analysis underscores the need for intellectual humility in the face of irreducible uncertainty. In messy situations we need mental maps to guide action. But our maps are incomplete representations of complex reality. We forget this at our own peril.

Tags: Frank Knight, profit, risk, uncertainty

November 14, 2009 12:44pm in book review, mental maps, turbulence | Comment