What if the world we live in truly is as unpredictable and contingent as the “noise” that surrounds us suggests?
In John Kay and Mervyn King’s interesting book it is a false sense of uncertainty that causes untold mistakes in politics, economics, and finance. Fundamental to their analysis is that there are different types of uncertainties. Economists used to distinguish between risk and uncertainty; Kay and King calls these resolvable uncertainty and radical uncertainty. Before defining these two concepts, we need to understand what Kay and King argue is at stake in this distinction. Mistaking radical uncertainty for manageable risk, they argue, caused the 2007-2008 financial crisis, and the subsequent decade of stagnant economic growth. Failing to recognize the distinction, they argue, has given the lie to every macroeconomic model yet invented. In politics, the same failure to recognize radical uncertainty led to the hubris of Robert McNamara’s “whiz kids,” and the debacle of escalation in Vietnam. It also helped bring about the Second Gulf War. On a more personal level, failure to recognize the difference has led people to adopt incorrect strategies when planning for retirement. Radical uncertainty is everywhere, they argue. And the modern inclination to conflate it with manageable uncertainty causes dramatic, repeated mistakes in decision-making.
Uncertainty Versus Risk
Largely mentioned only in passing today in the economics of information and uncertainty, in an older literature, “risk,” or manageable uncertainty, is what we could call a known unknown. That is, while the outcome is unknown, we know the probability distribution over the possible outcomes. A coin toss, for example, produces a head or a tail with the known probability of fifty-fifty.
In contrast, “uncertainty” in this older parlance is an unknown unknown. We don’t have a probability distribution over possible outcomes. Indeed, we might not recognize the possibility of an outcome at all. Kay and King’s “radical uncertainty” shares a lot with Nassim Taleb’s “black swan” events. An unanticipated invention, or a heretofore unanticipated viral threat, are “uncertainties.” It makes no sense to assign probabilities to something that we have not yet imagined might exist. And even if we know something exists, we may not yet have a clue as to the shape of the event’s probability distribution.
John Kay and Mervyn King seek to resurrect for day-to-day decision-making the significance of distinguishing between risk and uncertainty in Radical Uncertainty. They press the relevance and significance of the older distinction between decision-making under risk and decision-making under uncertainty. The thing is, decision-makers rarely have the option of not making a decision despite the possibility of radical uncertainty. Indeed, not making a decision is to make a decision. The upshot for Kay and King is the need to recognize the inescapability of applying practical wisdom given the real uncertainties faced in everyday life. In a sense they call for recognition of the implicit wisdom provided by intuition, experience and deliberation. Their argument also offers a brief in favor of the usefulness of minds trained in the liberal arts rather than simply in instrumentalist technical disciplines. This distinction between radical uncertainty and manageable uncertainty is a useful one that every day policymakers often recognize implicitly on their own, at least when not misdirected by the experts to fetishize mathematical and statistical models.
While their book was published just as the coronavirus pandemic first arrived in the West, decision-making in response to the pandemic neatly illustrates Kay and King’s argument.
Decision-making and policy-making in response to the pandemic occurred in an environment of radical uncertainty, they would argue, not in an environment of risk or manageable uncertainty. At least not at the start. Uncertainty—that is, unknown unknowns—existed, and still may continue to exist, at several levels. The first level of uncertainty pertained to the timing and nature of the pandemic itself. For decades now experts have warned policymakers of the near certainty of a global pandemic at some point in the future. While that sounds like a confident prediction—a probability approaching one—there was no basis to form beliefs regarding the probability of when the next pandemic would actually occur, nor the type of the disease.
Once the pandemic started, there remained (and perhaps remains) insufficient information to assess the course of the virus and, hence, to craft timely and appropriate responses. It is the lack of appreciation of this point—of the difference between decision-making under radical uncertainty relative to decision-making under known risks—that accounts for much of the confusion and dispute over how to respond to the virus. Decisions had to be made irrespective of the nature of the uncertainty. Better decisions would have been made on all sides, Kay and King would argue, had the radical uncertainty of the pandemic been recognized expressly rather than mistakenly translated into wildly mistaken statistical guesses by ostensible experts.
Most of Kay and King’s analysis circles around sketching the mistakes they view as resulting when situations of uncertainty are mistakenly framed as situations of risk. The 2007-2008 financial collapse and subsequent recession is the poster child for their argument; they recur to it repeatedly in the book. Additional examples span personal finance to policy making. They criticize “retirement calculators,” for example, for requiring that the user insert specific numbers for unknown-unknowns such as one’s expected lifespan, or one’s expenses in retirement. (Although they would presumably approve use of these calculators if people insert a range of values in an attempt to identify robust estimates of retirement needs.) Similarly, they criticize what they style as the quantitatively-charged hubris of Robert McNamara’s approach to defense and military decisions in the 1960s.
Models Are Not Entirely Useless
But Kay and King are careful to avoid throwing the baby out with the bathwater. While they criticize the reification of mathematical and statistical models, they do not reject them entirely. Mathematical models, they argue, can be helpful simplifications of reality. They can provide useful insights into behavior and outcomes. So, too, statistical models. What they reject is the fetishization of these models. They worry about the effects of processes by which analysts mistake mathematical simplifications of the world for the world itself, and are troubled by the ways that scientists ignore the inherently backward looking nature of statistical models in projecting their results into the future.
There certainly is a continuing need for modesty when applying modern social science tools to real life problems and policies. Kay and King’s argument can be taken as a call for recognition of the irreducible need for applying practical wisdom in day-to-day decision making, whether personal, social, economic, or political. Their argument can be taken to dovetail nicely with Thomas Sowell’s old argument distinguishing between “constrained” and “unconstrained” visions in his now largely ignored book A Conflict of Visions. Sowell’s “constrained vision” counsels against putting too much faith in abstract models over experience, and instead trusting “senses trained through experience to discern good and evil” (Heb 5.14). As I mentioned earlier, in a very importance sense, Kay and King’s argument can be considered a brief for the continuing and central significance education in the liberal arts and its role in developing prudential judgment.
And, yet, Kay and King seem to be of two minds on how to respond to uncertainty given the necessity of making decisions even when uncertainty prevails.
On the one hand, numerous times throughout the book, they suggest the only answer in the face of true uncertainty is for the analyst to say, “I don’t know.”
They write, for example, that “Different individuals and groups will make different assessments and arrive at different decisions, and often there will be no objectively right answer, either before or after the event” (emphasis added). Similarly, “[uncertainty entails] a world of uncertain futures and unpredictable consequences, about which there is necessary speculation and inevitable disagreement—disagreement which often will never be resolved” (emphasis added).
Yet that does not help when people cannot avoid making decisions under uncertainty. How much money should I be saving in my retirement accounts? “I don’t know.” Should our bank invest in this new financial instrument secured by real estate mortgages? “I don’t know.” Should we close the university in light of the coronavirus or keep it open? “I don’t know.” Can we safely keep non-essential businesses open as long as people wear masks, wash their hands, and distance themselves socially? “I don’t know.”
The problem is that this epistemological humility ends up undercutting Kay and King’s own criticism of mathematical and statistical models: How do they know unjustified reliance on mathematical and statistical models produced worse outcomes than ignoring them would have produced?
Hindsight is 20-20. Cherry-picking a list of decisions that turned to have been wrong retrospectively does not entail that we could have known they were mistaken in before the fact. Every losing bet is regretted after the fact. But if we do not know—cannot know—the decision counterfactual in advance, then judgment cannot be made. The examples of mistaken judgment accumulated in Kay and King’s book are nothing more than selecting the data that agrees with their hypothesis. Their own argument tells us that they cannot know that what they claim to know about mathematical and statistical models.
Kay and King seem to recognize this on some level. So they also provide a second, less pessimistic answer. An answer in which uncertainty can in fact be “successfully managed.” Yet exactly how this managing of radical uncertainty successfully occurs is never entirely clear. We get handwaving jargon that appeals to “narratives” and “legal reasoning” and evolutionary outcomes.
[W]e manage radical uncertainty in a context determined by the knowledge we have acquired through education and experience, and we make important decisions in conjunction with others—friends, family, colleagues and advisers.
Successful decision-making under uncertainty is a collaborative process.
Humans thrive in conditions of radical uncertainty when creative individuals can draw on collective intelligence, hone their ideas in communication with others, and operate in an environment which permits a stable reference narrative.
Within the context of a secure reference narrative, uncertainty is to be welcomed rather than feared.
Consider just the notion included in these last two quotations. That of “stable” or “secure reference narratives.” On the one hand they argue that successful management of uncertainty requires these stable and secure reference narratives. Yet these “reference narratives” apparently cannot not become too stable and secure, because “bureaucracies staffed by risk-averse individuals determined to protect their personal reference narratives” is also “associated with paralysis of decision-making.” In his book, The Black Swan, Nassim Taleb is also critical of the ability of people mistakenly to commit to false narratives as well as to false mathematical and statistical models.
The solution then to managing uncertainty for Kay and King? To recognize “radical uncertainty and adopt policies and strategies that will be robust to many alternative futures . . .” This would seem to be a form of optimal behavior in the face of uncertainty, a form their analysis rules out.
This line of analysis ultimately leave us with bromides like this: “[I]f we control risk we can not only manage but positively enjoy uncertainty.” Really? I somehow doubt many policymakers enjoyed the necessity of planning for a pandemic under uncertainty. It is unwarrantedly asymmetrical for Kay and King to emphasize the positive and enjoyable side of decision-making in an environment of radical uncertainty. It seems just as easy for humans to respond to radical uncertainty in dangerous and fateful ways. There is no reason to paint a smile on the phenomenon.
Remember, however, this is a second answer Kay and King provide, one less pessimistic than their first answer. That Kay and King ultimately succumb to the temptation they criticize in others does not mean that the centerpiece of their argument—the distinction between risk and uncertainty—is mistaken.
Political Institutions as a Response to Radical Uncertainty
The distinction between radical uncertainty and manageable risk is both relevant and of practical value in everyday life, even by those who do not recognize the technical distinction. Consider for example the central role radical uncertainty plays in the Illinois Supreme Court’s justification of the broad delegation of legislative authority to public health boards in its oft cited 1922 decision in Barmore v. Robertson:
The necessity of delegating to an administrative body . . . is apparent to everyone who has followed recent events. Legislatures cannot anticipate all the contagious and infectious diseases that may break out in a community, and to limit the activities of the health authorities to those diseases named by the legislature in the act creating the administrative body would oftentimes endanger the health and the lives of the people. There is probably not a legislature in the country that would have named the deadly Spanish influenza as a contagious and infectious disease prior to the epidemic of that disease that took a greater toll of lives throughout the country than any other epidemic known in this country. In emergencies of this character it is indispensable to the preservation of public health that some administrative body should be clothed with authority to make adequate rules which have the force of law and to put these rules and regulations into effect promptly.
To be sure, bureaus and agencies are subject to the same temptations that we all face, including the possibility of making decisions based on too little information or overly reified “reference narratives” (in Kay and King’s jargon). Nonetheless, the Illinois Supreme Court makes a straightforward appeal to what Kay and King call radical uncertainty in justifying a broad delegation of legislative power to an executive branch agency. It is an argument that needs to be dealt with whether one approves the delegation or not.
More to the point, when a state court over a century ago intuitively, if implicitly, draws on the concept of radical uncertainty to articulate the need for an institution to deal with that phenomenon, it is clear that Kay and King are wrestling with something real, and so need we. It is an important concept for modern life, notwithstanding the need Kay and King seem to feel to minimize the epistemologically pessimistic implications of their argument.