Journal Article
No. 2016-3 | January 28, 2016
Uncertainty: A Diagrammatic Treatment


The purpose of this paper is to clarify the difference between the mainstream and Keynesian understandings of uncertainty which persists in spite of superficial similarities. It is argued that the difference stems from the mainstream habit of thinking in terms of a full-information benchmark, where uncertainty arises from incomplete information. Degrees of uncertainty (or ambiguity) refer to the quantifiable extent of incompleteness. In contrast, Keynesian uncertainty cannot, even in principle, be eliminated. By treating uncertain knowledge as the norm, Keynesian uncertainty theory analyses differing degrees of uncertainty in relation to grounds for belief and thus considers the cognitive role of institutions and conventions in influencing the degree of uncertainty. The paper offers a simple diagrammatic representation of these differences, and illustrates its use with different depictions of the crisis, its aftermath and the policy response appropriate to each understanding.

JEL Classification:

B41, B5, E00, G01


  • Downloads: 1147 (Discussion Paper: 1200)


Cite As

Sheila Dow (2016). Uncertainty: A Diagrammatic Treatment. Economics: The Open-Access, Open-Assessment E-Journal, 10 (2016-3): 1–25.

Comments and Questions

Romar Correa - The paper by Sheila Dow
January 29, 2016 - 16:51

We remain grateful to Professor Dow for her insights and subtlety. I was reminded of the distinction made by Kenneth Boulding, an early advocate of open systems reasoning in economics, between data and information. The category beloved of the mainstream is, properly, data. Information is data in a context. The ...[more]

... latter is the arena of surprise and novelty.
None of this might matter for understanding the crisis. Charlatans and robbers abounded. Look forward to more of Sheila Dow on ethics here!

Peter Smith - Sheila Dow, paper on Uncertainty: A Diagrammatic Treatment.
February 14, 2016 - 14:42

Comments on Sheila Dow, paper on Uncertainty: A Diagrammatic Treatment.

[p3] Agree with her positive evaluation of Keynes in this matter, but not sure what ‘philosophically-grounded’ means, given the tendency in Western philosophy to try for an a priori basis that supports a prejudged conclusion popular with one’s own ...[more]

... constituency (look at, say, Rawls versus Nozick). I think Keynes simply thought about the problem, rather than choosing an ultra-simplified model that was consistent with a seemingly rigorous case for laissez-faire.

[p4, SEU] Not sure how far this comment is supported by current research, but my experience as a consultant and trainer back when SEU was a hot topic strongly suggests that, when practitioners give a component of some course of action that they lean towards a high SEU, they are often evaluating their ability to deal with the consequences of choosing it, not evaluating probability at all. Anyone know anything about this?

[p5, quantitative probabilities] There are problems – ignored or unrecognised – even within the traditional approaches that conflate risk and uncertainty. Many economic time series actually have strongly non-Gaussian probability distributions; these throw up far more extreme events than the so-called Normal distribution predicts. Worse, it requires a huge amount of data to decide whether this is the situation we are in. Mandelbrot and Hudson’s The (Mis)Behaviour of Markets gives a good non-specialist introduction. They showed that the Dow (1915 to 2000) is a good example of a non-Normal dataset. Modern financial theory mostly still assumes that ‘Normality’ rules; with that assumption, the probability of the wildest observation of the index is about 10 to the power of minus 50. (To give an idea of the scale of such a factor, multiplying the size of single, medium-sized atom (say, of iron) by 10 to the power 50 would produce an object with a diameter larger than that of the known Universe.)
This points to an issue that is tangled up with the urgent problems of economics, and the way that theory, policy, and practice are entangled. There is a significant body of evidence on this issue; this suggests that, even where we are sure we have a quantifiable degree of variability, actually quantifying it may be an intractable problem. This fact has been suppressed or wilfully ignored by mainstream economists for decades (see Mandelbrot and Hudson, op cit) – yet it has a strong influence on our expectations, however we translate those into concrete choices.

[p9, ambiguity] Not sure that this is the most potent definition of ambiguity: often we do not know what a particular signal (e.g., from the markets) means, which (I think) is a separate issue.

[p11, “In Metaphysics, in Science, and in Conduct …”] It’s been said before, but Aristotle got there first, in the Nichomachean Ethics:
... in discussing [these ethical] subjects, and arguing from evidence, we must be satisfied with a broad outline of the truth; that is, in arguing about what is for the most part so, from premises which are for the most part true, we must be content to draw conclusions that are similarly qualified ...

[p12, ‘human’ logic] I was surprised not to see fuzzy logic introduced here, as it is the formalized version of our everyday reasoning. It is also the more general version, of which standard logics (i.e., those that allow only two truth-values, viz, ‘true’ and ‘false’) apply only to the special – and rare – cases where we have pathological levels of certainty.

[p12, Keynesian weight as an ordinal concept] [p13 numerical limits] I have suggested elsewhere than rational decision-making under uncertainty does not depend upon ranking the options – as, indeed, Prof Dow is careful not to endorse. The alternative I put forward grows out of Ian Mitroff’s idea of problem structure, problems being well-structured, ill-structured, or wickedly-structured, according to where they lie on the continuum that runs from certainty or mere risk; through radical uncertainty; to a politicised arena in which values and paradigms differ among the actors. At risk of citing myself (no, it isn’t, it is the certainty of doing so!), this is set out in Smith, PJ (2011) The Reform of Economics, Chapter 10 – I’m not claiming any deep originality for that material, but the reference does pull together some very scattered literature.

[p15, and elsewhere, use of single quotes around terms (e.g., the word ‘ambiguity’), to distinguish a modified term from its parent] While the author’s intent is clear (and is properly defined at first usage), it would be kinder to the reader to have introduced a distinct, modified label. Relying on punctuation marks to distinguish two related terms looks like a recipe for confusion.

I hope this is useful.
Peter Smith.
Mail: Springcott, Chittlehamholt, Devon EX37 9PD.