A New Typology of Uncertainty (for Decision-Making)

Abstract

Argues that all past typologies are flawed, if not wholly wrong. Explains and argues a new primary typology of uncertainty based on treatability. Explains and argues a new (supplementary) secondary typology based on the elements of the decision-making process. Describes a (supplementary) tertiary typology based on a 5W1H perspective.

Appendices

Supplement on Knightian Uncertainty Issues

Knight’s (1921) seminal work on uncertainty has reverberated for over a century, especially in the entrepreneurship field. However, one hundred years of thinking about its ideas (and testing some of them) has exposed some issues worth considering now.

Knight’s (1921) work has had a large impact on entrepreneurship research. Even a century later, his conceptualization of uncertainty continues to influence work on the role of the entrepreneur in bearing specific informational challenges (e.g., Townsend et al., 2020). His delineation of risk from non-risk uncertainty—where the former entails problems that include the distributional information necessary to compute expected outcomes while the latter does not—has resulted in the latter being termed Knightian uncertainty. In his model, only under this kind of uncertainty can entrepreneurial profits arise—realized by bearing such uninsurable decisions through commitments and actions in pursuit of the profits that come to the residual holders of the contracts that constitute their ventures.

Knight’s Model as a Theory of Rents

Special for its time, Knight’s (1921) model can now be considered as just another explanation of super-normal rents. Knight’s super-normal profits are the result of revenues that are inflated by an undersupplied product combined with input costs that are reduced by their under-demand—with such inflation and reduction arising because of the low number of entrepreneurs-as-producers in that specific (new) product market. (Given that they constitute the few special people who are willing to bear uncertainty, they have little competition in selling their goods on one hand and buying their inputs on the other.) What differentiates Knight’s model is the story about what gives an entrepreneur the ability to produce the valued-but-undersupplied products—her unique judgment of what to do (i.e., which product to make, how, and in what quantity) when faced with non-risk uncertainty. (Luck, confidence, and the ability to execute also help.) In Knight’s model, the entrepreneur is the residual holder of a set of contracts because she alone is willing to bear the non-risk uncertainty as the leader of her venture that her employees have contracted with.⁸ (Such a private arrangement also limits any spillover of her unique decisions based on her valuable judgment.) That story of rent-earning is no longer unique—as it simply presents just another version of the narrative where someone with a valuable, rare, inimitable, and appropriable capability can use it to gain an advantage (e.g., see the capability, resource, and knowledge-based views—e.g., Barney, 2002; Grant, 1996).

Although that kind of rent-logic is no longer special, the connections Knight’s (1921) version has with entrepreneurship, with strategic decision-making, and with uncertainty have been of continued interest, spiking again recently. Its connection with the entrepreneurship field remains strong because that field continues to look for defining theories, especially those that provide solid micro-economic arguments and that have survived decades; so it is not surprising it continues to be an easy and popular citation in addition to remaining an inspiration for new theorizing (e.g., Rindova & Courtney, 2020). Its connection with strategic decision-making remains strong, as managers struggle to deal with VUCA contexts where unknown unknowns and unimaginables make traditional tools based on scenarios, adaptability and options less applicable (Arend, 2020). And, its connection with uncertainty also remains strong, especially recently, as a renewed interest in incomplete informational problems involving unknowns has occurred. That renewed interest is a result of the historic wave of papers that had unfortunately replaced real uncertainty with risk (e.g., through subjective expected utility, beliefs and learning—Hodgson, 2011) ‘hitting a (predictable theory-related) wall’ when strategic decision-makers increasingly found themselves confronting uncertainty-related problems without any solid basis for the justified priors required for such expected-returns-based optimizations. And, so, the subsequent gold-rush to describe possible solutions to these unoptimizable problems was on. But, danger now looms when such follow-on work builds upon Knight’s model, especially when that model is loosely interpreted. These underappreciated dangers to academics and to practitioners are what we focus on here.

The Three Dangers of the Knightian Model

Knight’s (1921) model of entrepreneurial activity (one based on the market failure of incomplete information) involves three main dangers—one involving the main premise of the model, one involving the main implication of the model, and one involving the main definition of the model. We explain why those dangers can matter materially to academia and to practice after describing each.

The Premise Danger

For Knight’s (1921) story to work, there must exist, for all relevant parties, a minimum level of uncertainty that is beyond risky (i.e., without an expected value calculation, and so uninsurable). Essentially, if even one party confronts only risk in their decision, then that party could (profitably) provide insurance to others. And, when that occurs, no interesting party faces non-risk uncertainty and there is no entrepreneurial activity—as Knight would define it—required. No special judgment is needed. It is business-as-usual, as the downside is insured.

To illustrate, consider a one-factor production decision. It is a simple optimization problem with one input. The objective for the decision-maker is to maximize the profit on her productive effort. The benefit is accounted for in terms of revenue constituted from the output of the effort (e) multiplied by the output’s price (p). The variable cost is accounted for in terms of the square of the effort level exerted. And, there is a fixed investment (F) to enter that market, such that the problem is:

$$MAX \pi \left(e\right)=p\bullet e-{e}^{2}-F, p\ge 0, F>0$$

The optimal effort is a function of the price level (e* = p/2). There is a participation condition (assuming the opportunity cost is normalized to zero) of: $p>2\sqrt{F}$. So, when the non-risk uncertainty concerns the value of p, then there is no optimization possible. It would then be unlikely that the decision-maker enters this market (i.e., when the value of p is unknowable, as well as its distribution, and the distribution of that distribution, and so on) because there is no guarantee or justified expectation of any positive gain from entering (e.g., given that the price may well lie below the participation condition). If, however, one of the relevant parties has confidence (i.e., subjectively perceives, regardless of any objective facts) that the expected price level lies above the participation condition, not only would she rationally enter the market, but she would also offer others the minimum participation price (plus an epsilon-sized bonus just above zero) to buy their product. Doing so essentially offers them insurance while also profiting from the expected difference in price without signaling what that level is (so it could not be calculated by others). In that case, she essentially ensures that no Knightian entrepreneurial activity occurs. (Note that we are not stating that she is using judgment while she correctly perceives the same level of uncertainty everyone else does, in which case she may be the sole Knightian entrepreneur here; what we are instead saying that she perceives a lower level of uncertainty than others and so does not need to apply any unusual judgment.)

Thus, this premise of a universality of a minimum uncertainty level (i.e., where that uncertainty is a level beyond risky) is dangerous because of its sensitivity to even a minor violation. It severely weakens the robustness of one the most famous theories explaining why entrepreneurs exist. From a theoretical perspective, while it is not unusual to assume that most factors are homogeneous across relevant parties (e.g., in order to simplify a model and focus its analysis), here there is more at stake because of the discontinuity that occurs when even one party differs. It is also troubling that the minimum necessary uncertainty level is singled out for homogeneity when so many factors related to it—like the judgment, confidence, and execution for those facing it—are all assumed heterogeneous. It is troubling because judgment—and the confidence in it needed to act—should be linked to the level of uncertainty being perceived. Reducing perceived uncertainty should increase the confidence a decision-maker has in her judgments, especially any that are unique. That link is embodied in experience. Heterogeneity in decision-maker judgment and confidence is based on past experience (Likierman, 2020), specifically here, on experiences with whatever level of uncertainty had been perceived. So, if experience not only links past perception to the current perception of uncertainty but also the current perception of uncertainty to judgment and its confidence, then the idea that differences in judgment can occur without related differences in perception seems unlikely. With that unlikeliness comes the violation of the focal premise of the model. With that unlikeliness also comes the question of whether the model can be tested (i.e., whether it is possible to separate differences in judgment about uncertainty from differences in perception of that uncertainty). And, with that question of premise and testing, comes the suspicion that such a model’s prescriptions may be quite weak in the real world.

Further, when the premise is wrong, we get a mislabeling of what Knightian entrepreneurship is. That mislabeling then poses legitimacy dangers for the original model and any model that builds upon it. For academia, holding the original model’s premise here means an under-appreciation in our theories of the sensitivity to uncertainty-level perception involved. It may also mean too few studies of decision-making and opportunity-exploiting based on that uncertainty-perception variance that Knight did not appear to consider.

Any practitioner misinterpreting Knight’s premise also faces dangers. There would be an under-appreciation of the variance in how people perceive the same event (e.g., Sjöberg, 2000), including in how they delineate risk from beyond risk, and that could lead to mistakes in how the practitioner predicts how rivals, customers and others will react to uncertainty-related problems. When the practitioner perceives non-risk uncertainty while others perceive only risk, she will miss out on using those others to insure what she believes is uninsurable (when she follows Knight). On the other hand, when she perceives only risk while others perceive non-risk uncertainty, she is likely to be exploited as an insurer for those others (when she follows Knight). In either case, following Knight’s premise may be very costly.

The Implication Danger

Knight’s (1921) story links non-risk uncertainty directly to entrepreneurial activity, doing so with the primary logical implication that if some non-risk uncertainty is necessary to induce entrepreneurial activity, then more of it will induce greater activity. The more of it refers to both quantity (i.e., more decisions or markets involving non-risk uncertainty) and quality (i.e., higher levels of that uncertainty). Similarly, the greater activity refers to both greater quantity (i.e., more entrepreneurs because more markets have the necessary uncertainty level or because any one market involves more uncertainties than one venture can exploit) and quality (i.e., better performance for the entrepreneur in markets that are providing wider or deeper uncertainty-related opportunities to exploit). This implication is based in the text of Knight’s (1921) book: he speaks to the larger profits that can be made when greater uncertainties exist in the markets that lack rivals with similar uncertainty-bearing capabilities (p. 230) as the: ‘…importance of uncertainty as a factor interfering with the perfect workings of competition…’. He also speaks to diminishing positive returns to entrepreneurship as being a function of uncertainty: (p. 286), as: ‘The question of diminishing returns from entrepreneurship is really a matter of the amount of uncertainty present’. (Note that diminishing returns can only occur when returns were positive and at levels high enough to measure material reductions.) And, prior to this, he commits to the linkage of uncertainty to income (p. 232) as: ‘this true uncertainty … accounts for the peculiar income of the entrepreneur’.

The danger arising from this implication is that we know uncertainty has costs, even for Knightian entrepreneurs. Knight (1921) even acknowledges—in a brief concession—that some of his entrepreneurs do fail because their judgment is wrong, their confidence is misplaced, or they are just unlucky. But, the most detailed and common Knightian story does not sufficiently reflect upon those failures, nor does it explicitly account for the other negative effects on the windfall-like incomes of the initially luckier entrepreneurs. It is inaccurate, theoretically, to concentrate a model on only the good side of a factor, especially when (even in 1921) the bad sides were known, or at least suspected. For example, it was and remains known that uncertainties have negative effects on even the performance of a Knightian entrepreneur (e.g., by suppressing demand from uncertainty-averse consumers). Not acknowledging that the kind of positive logical relationship that Knight implies between uncertainty and entrepreneurial outcomes is actually likely inaccurate (because it is incomplete) is dangerous. It is dangerous because holding to such relationships ensures that entrepreneurial activity—in terms of level and performance—will not be properly understood in terms of its complex association with uncertainty (e.g., in the field and in empirical studies). We may miss out on studying the interdependencies of the positive and negative effects of uncertainties, perhaps moderated by who the entrepreneur is, and which opportunities are involved. We may miss out on better specifying when entrepreneurial activity should actually arise and what policies there can be that decrease the linked negative effects of uncertainty. At the very least, any theoretical work that builds upon such a questionable implication stands on shaky ground and may damage the legitimacy of the fields it is contributing to.

The practical dangers arising from the overly-positive implication of Knight are even more stark. We know that real-world behaviors are adversely affected by uncertainty. For example, experiments have proven that risk-aversion and ambiguity-aversion are real and involve significant premia that reduce the efficiency of economic transactions and markets (e.g., Becker & Brownson, 1964). We have even seen the devastation that the multi-dimensionality of uncertainty linked to the recent Covid pandemic has had on the newer, smaller—often considered more entrepreneurial—ventures compared to more diversified, or more traditional corporations (e.g., Cui et al., 2023). In this light, the practical implication of Knight’s linkage of uncertainty with entrepreneurial activity and performance seems naïve (if not myopic) at best, and unfortunately dangerous at worst.⁹ Interpreting Knight as truth about how uncertainty is linked to entrepreneurship will reduce the likelihood of better uncertainty-related opportunity outcomes in the field, and any policy based on those relationships will not work as expected (likely under-performing because it does not account for important negative effects).

The Definition Danger

Knight’s (1921) story is famous for its delineation of risk from uncertainty; however, its actual definition of that non-risk uncertainty is problematic due to its lack of crispness (e.g., Spender, 2006). The under-definition leads to multiple dangers: First, there are dangers that arise because there are actually many types of non-risk uncertainty that meet Knight’s definition (i.e., of being uninsurable, unpredictable, and unoptimizable).¹⁰ But, only some of these types matter; and so, if these are confused with ones that do not matter, then damage is likely. For example, even under non-risk uncertainty some problems may only be trivially non-optimizable (e.g., when the unknown’s distribution is bounded by a narrow, but still profitable, range) while others can pose an existential threat (e.g., because the range of payoffs may include bankruptcy-level losses). Second, there are dangers that arise because the nature of the unknowableness of the focal uncertain factor lacks critical clarity. For example, when the unknowable is confused with the unknown-but-knowable, it can jeopardize venture performance (i.e., in terms of missing out on uncertainty reduction or spending too much on trying to reduce the irreducible).

The lack of clarity arises in Knight’s (1921) work because he mixes theoretical and practical concerns regarding how unknowableness can be dealt with. In theory, non-risk uncertainty is meant to be borne as is by entrepreneurs who have (or believe they have) better judgment when facing what they perceive is unknowable. That unknowableness must be irreducible prior to when the strategic decision must be made (Ramoglou, 2021) in Knight’s model. And, when the decision is right as that uncertainty is borne, the entrepreneur enjoys super-normal rents. However, and unfortunately, Knight also describes five other ways to deal with non-risk uncertainty where each of those ways transforms the ex ante unknown it into something (more) insurable or optimizable. These are practical approaches where the uncertainty is not borne, but rather is addressed through standard operating procedures where the expected outcome is in the form of normal rents.

The dangers to the field arising from the under-definition of Knight’s uncertainty are twofold: First, without a proper typology of the possible non-risk uncertainties, the term Knightian uncertainty involves too coarse a basis from which to build new theory. Without the provision of the full problem context—including the specific variant of non-risk uncertainty—any potential approaches to non-optimizable problems cannot be properly analyzed, let alone prescribed, with any legitimacy. For example, approaches for handling choice-option uncertainty differs from handling outcome-set uncertainty (e.g., Packard et al., 2017). Such differences in unknowns related to choices, outputs, probabilities and payoffs matter. For example, one unknown but bounded probability can often be dealt with in a decision; but, one unknown outcome or one (relatively unbounded) payoff that could occur for any choice theoretically cannot be dealt with. Second, without a clear delineation of the character of the unknowableness involved—specifically whether it is unknown-but-knowable or not—the confusion over what Knightian entrepreneurs actually are (i.e., those who bear the unknowable unknowns) will continue. They will be mistaken for entrepreneurs that, as part of their standard operating procedures, are only better at reducing a given unknown to something more known. That leads to mis-categorization, and the subsequent mistaken theorizing and testing that comes from it (e.g., to the point of making mistakes in prescriptions, investments, and empirical investigations).

The practical dangers arising from the under-definition are also twofold: The primary practical danger results from a lack of a proper typology of non-risk uncertainties. Without that typology, it is not possible for practitioners to study (in the field) which heuristics are effective against each type, nor is it possible to provide type-specific prescriptions from data, experience or theory to apply in the field. The second danger is that not having a clear delineation of unknowability is likely to result in mistakes involving treating unknowables as knowables and vice versa. In the real world, the former is likely to involve decision-makers trying to seek more information (e.g., through experimentation with the market), while the latter is likely to involve avoidance of the problem, or waiting it out, or retaining flexibility to respond to the outcome; with the respective dangers being in the forms of wasted effort and overconfidence, and of a wasted opportunity to gather the information to make a better proactive decision.

Discussion

Misinterpreting Knight’s (1921) model poses dangers: It should not be interpreted as a robust theory of entrepreneurship or of strategic decision-making under uncertainty, but instead as a model based on a knife’s edge premise over the shared minimum level of non-risk uncertainty perceived. It should not be interpreted as describing a simple positive relationship between non-risk uncertainty and entrepreneurial activity (or performance) because, in both the academic and practical worlds, uncertainty’s effects are complex and most often negative (Taleb, 2012). And, it should not be interpreted as clearly delineating all non-risk uncertainties in a useful manner, given there are many types of those uncertainties that should be addressed differently. All together, such potential dangers weaken Knight’s contribution to today’s understanding of entrepreneurial and strategically uncertain phenomena; but, that is not properly recognized by many who continue to cite it and build upon it. Such misinterpretations endanger both precise modeling and in-the-field testing, in addition to the subsequent practical prescriptions that could arise from each.

Reconsidering the premise danger, we know that a universally perceived minimum level of non-risk uncertainty for any focal problem is unlikely in the real world. We know from research that risk perception varies among business people (e.g., Mitchell, 1995). Similarly, we also know that uncertainty perception varies significantly across real decision-makers, including entrepreneurs, with differences in perceived uncertainty levels correlated with many factors, including those associated with the individual, the organization, and the community (e.g., Chawla et al., 2012; Downey & Slocum, 1975; Elenkov, 1997). Such findings are likely to extend to the perception of any critical level of non-risk uncertainty, ruining the necessary condition for Knightian entrepreneurship to exist. However, such heterogeneity in uncertainty perception could be exploited by policy-makers by making it easier for the relevant insurance-type markets to exist, so that those seeing less uncertainty can reduce it for those who see more, to the advantage of all. By adhering to Knight, however, that won’t be done. Nor will studying what it means to bear perceived uncertainty (e.g., psychologically) so that those who do so can be better understood and compensated (Dimov, 2018).

Reconsidering the implication dangers, we know that increases in the quality and quantity of real non-risk uncertainty are likely to work against real venture performance regardless of the potential for new opportunities it creates. For example, a high level of perceived non-risk uncertainty will delay if not decrease demand. Such effects strain the cash flows of most ventures. Further, widespread non-risk uncertainty caused by such events as the pandemic do more than cause decision-making hesitation and cash-flow anxiety to small business entrepreneurs; they can lead to strategic mistakes (Knight, 1921), ones that are existential in such contexts, and understood as such by entrepreneurs. Based on the psychological and behavioral economics research (e.g., Ellsberg, 1961; Kahneman & Tversky, 1979), common forms of uncertainty aversion will cause real entrepreneurs will pull back from investments, reducing even the future performance of their ventures. Added to those reactions, there are often regulatory restrictions on operations, hours, and on capacity, that all harm rather than help especially the new and small ventures. The negative consequences of increased uncertainty—that Knight fails to properly account for—squarely implies there exists an important role for government in reducing the perception, if not the real effects, of widespread uncertainties faced by the (entrepreneurial) small business sector. Real policy should not rely on Knight (1921).

Reconsidering the definitional dangers, we know that it is unlikely that real decision-makers delineate among the various types of non-risk uncertainty, let alone whether they treat each type differently. Questions over who should be responsible for perceiving the uncertainty level or type, how the choice is made whether it needs to be dealt with, and what the sequence is of how to do so, all need to be explored. In the real world, Knight’s model has little to say in answering those questions. In following it, we miss out on important avenues for research, for practice and for policy-making.

The implication is that policy-makers have a mandate—especially post-Covid—to fund such research, first in lab studies (where the types and amounts of uncertainties can be better controlled to see if those make a difference to perceptions and behaviors) and second, in the field (to see how those differences translate into reality).¹¹ That research should provide a better basis for hel** economic actors to perceive uncertainty, and then to build better toolkits and expertise for applying those tools when Knight’s non-risk uncertainties confront them. Rather than relying on simple interpretations of Knight (and his version of uncertainty), we as scholars of entrepreneurship and strategic management and decision-making need to confront the limitations and dangers of that model and move forward, and we hope that this supplement has provided a basis for so doing.