Wednesday, December 8, 2010

Risk and Uncertainty in Post Keynesian Economics

The distinction between risk and uncertainty is fundamental in Post Keynesian economics, as it was in the economic thinking of John Maynard Keynes. While risk can be quantified, uncertainty simply cannot be quantified.

The notion of future uncertainty is partly related to David Hume’s difficult philosophical problem of induction. According to Hume, we have no basis for believing that induction is rational. Human beings use induction out of habit, but there is no rational justification for this belief in induction. Although a full discussion of the problem of induction is not what I want to do here, a few comments can be made.

In brief, induction is the use of a finite number of specific observations to make a general conclusion. The most common form is the use of past observations of events to infer a general conclusion about how nature will behave in the future (e.g., gravity will continue to operate tomorrow or 10 minutes from now). But we can just as easily appeal to observed events in the recent past or present to make a general conclusion about the past (e.g., miracles do not happen in the past). The conclusion of an inductive argument is only probable and never certain.

The trouble is that observation of a finite number of events in the past does not seem to justify the belief that events will continue to happen in this way in the future. One solution to the problem is to assume the uniformity of nature. But even here the assumption that nature is uniform requires an inductive argument, so the reasoning is circular (and note that the radical sense of uncertainty that emerges in the absence of a justification for induction and the uniformity of nature is rather different from Keynesian uncertainty, which relates to future events for which no measurable probability can be given, as we will see below). Various philosophers have tried to solve the problem of induction (Will 1953; Edwards 1965; Strawson 1952; BonJour 1998; Salmon 1974), but many believe there is no satisfactory justification. Hume concluded that we use induction out of habit but with no legitimate basis, and modern evolutionary psychologists might argue that our ability to instinctively reason inductively is an innate property of the brain (like language or moral sentiments), which we have acquired by evolution, even though there is no rational justification for it.

Probably the best solution to the problem of induction (for both the natural and social sciences) is the use of Karl Popper’s critical rationalism (even though Popper’s critical rationalism seems to be widely criticised in modern analytic philosophy [Musgrave 2004: 16–17], and another complaint is that, while many working scientists claim to be Popperians in method, in practice they behave like good Bayesian inductivists [Evans 2007: 33]).

Popper adopts the hypothetico-deductive method, with falsification (not verification) of hypotheses by empirical evidence the key. In hypothetico-deduction, we use the method of forming a hypothesis, deducing predictions or conclusions from it, then empirically testing the predictions or conclusions, and thereby attempting to falsify the hypothesis. Popper comes to an astonishing conclusion:
“I go further than Hume: I hold that inductive procedures simply do not exist (not even low-level ones) and that the story of their existence is a myth” (Popper 1983: 118; see also Musgrave 2004: 18).
Induction is also unnecessary: we simply do not need it. In the natural sciences and other scientific inquiry, what really happens is that we use deduction, especially by means of the modus tollens and the elimination of erroneous hypotheses by empirical evidence, in a process of trial and error (more on this below).

John Maynard Keynes’ Treatise on Probability (1921) was an attempt to answer Hume on the problem of induction. Keynes took the view that induction was justified by a relation of probability which was objective (although later he conceded that probability was not an objective relation). But Keynes also came to argue that there must be a distinction between risk and uncertainty.

Risk is something where a measurable probability can be given to outcomes, e.g., the probability of rolling a 3 when throwing a dice is 1 in 6 (Glickman 2003: 366). In contrast, we face fundamental uncertainty about many other events, and no measurable probabilities can be assigned (e.g., what the interest rate will be in 10 years time). This distinction between risk and uncertainty was also made by Frank Knight.

Keynes’s conception of uncertainty applies to what is technically called a nonergodic stochastic system (Davidson 2002: 187). Our economies are such complex systems, as are many other economic phenomena (e.g., stock markets and financial markets).

We face fundamental ontological or metaphysical uncertainty about many events in the future, a state of affairs which – when applied to economics – can be called “Keynesian uncertainty” (or uncertainty in the sense of Frank Knight, Keynes, George L. S. Shackle and Ludwig Lachmann).

Since there is fundamental uncertainty about many economic variables in the future (particularly in the long-term future), investment decisions cannot be based on a firm calculation of probabilities of future earnings. Since subjective preferences can change in the future, not all investment decisions today will be profitable in the future. The process involved in decision-making about investment and the action of investment itself is not rational calculation.

There is a time difference between production of commodities and the successful sale of those commodities at a profit, and so even the production of commodities has a speculative element, in which decisions to invest in production involve habits of minds, instincts, and conventions.

Thus the investment decisions of a firm are based on expectations of future earnings, which are not necessarily rational at all. When investment decisions are made, they are done under conditions of subjective expectations by business, and the expectations depend on what Keynes called “animal spirits” (for the original concept, see Gerrard 1994). Keynes discussed the factors influencing long-run expectations in Chapter 12 of the General Theory. Because of uncertainty about the future, expectations in investment decisions is not a matter of mathematical calculation. Decisions to invest are taken
“as a result of animal spirits – of a spontaneous urge to action rather than inaction, and not as the outcome of a weighted average of quantitative benefits multiplied by quantitative probabilities” (Keynes 2008 [1936]: 144).
The term “animal spirits” was borrowed by Keynes from Descartes (Gerrard 1994: 15), and whether or not modern psychology provides support for Keynes’ idea that we have a “spontaneous urge to action rather than inaction” is really irrelevant. The fundamental point here is that, because of uncertainty about the future and changing subjective preferences of consumers and other exogenous factors driving supply and demand, there can be no genuine rationality in expectations. Expectations are subjective, and the investment decision is essentially non-rational.

There has been a resurgence of interest in subjective expectations since the great recession of 2008/2009, and most notably the New Keynesians George A. Akerlof and Robert J. Shiller have published a book studying this subject (Animal Spirits: How Human Psychology Drives the Economy, and Why It Matters for Global Capitalism, 2009).

Neoclassical economics in its various forms (e.g., general equilibrium analysis, the efficient market hypothesis, and New Classical economics) assumes the ergodic axiom, the belief that the future can be known or events in the future given objective probabilities. But recognition of the ontological uncertainty of events in the future and changing subjective preferences means that events cannot be quantified in terms of probabilities. This destroys the basis of rational expectations and other neoclassical theories.

In the face of the shock of the financial crisis in 2008, the severe global recession in 2008–2009, and debt deflation, business expectations across the world plummeted. The destruction of optimistic expectations is undoubtedly an important factor that will mire many economies around the world in periods of stagnation with low growth and high unemployment in years to come. The Keynesian resurgence and stimulus packages in many countries in 2008–2009 prevented a depression, but much more stimulus is needed in the major economies to restore better growth and to bring unemployment down (there are of course other problems too, such as poorly regulated financial markets and loss of manufacturing in countries like the US that will have to be fixed). Government spending and fiscal stimulus can help to overcome the poor business expectations in many countries.

To return to an earlier point above, how can we justify government intervention in the face of uncertainty? If one accepts that induction can be justified (perhaps along the lines of Will 1953 or Strawson 1952), then obviously one could use an inductive argument for government intervention.

But, if we argue that induction has severe philosophical problems and cannot be justified, how in fact do we justify the hypothesis that government stimulus packages should be used and will work? My answer is that we do not need inductive arguments. We can use Karl Popper’s hypothetico-deductive method to test Keynesian hypotheses about stimulus. Our Keynesian models of stimulus will make predictions about what will happen to an economy, and, if they are not falsified by the empirical evidence, they will have passed the test of falsification, just as they have many times in the past. Popper argued that the hypothetico-deductive method was fundamental to both the natural sciences and social sciences, and he makes a good case for this (Popper 1976: 130–143; Milonakis and Fine 2009: 262–263). In short, induction is unnecessary, and economic methodology can be hypothetico-deductive with falsification of hypotheses by empirical evidence, as M. Blaug (1992) has also argued.

The issue of how Popper’s critical rationalism should be properly applied to economics is not my purpose, and the issue is a difficult one.

Of course, there are numerous economists who do not think that a strict Popperian methodology of economics is workable (Caldwell 1985: 126), and on this subject one can consult as a starting point the critical essays in N. De Marchi (ed.). The Popperian Legacy in Economics: Papers Presented at a Symposium in Amsterdam, December 1985 (Cambridge and New York, 1988), with a response by L. A. Boland (1990–1992).

Lawrence A. Boland contends that economists promoting a Popperian methodology for economics have not fully understood Popper’s critical rationalism, and have been misled by Imre Lakatos’s alleged distortion of Popper’s thought by overemphasizing the role of falsificationism (Boland 2006: 222–223). Boland (2006: 223–224) believes that an Imre Lakatos-inspired and pseudo-Popperian method has been adopted by some proponents of a Popperian methodology for economics.

Although this question requires a post in its own right, there is no doubt that Popper argued for the unity of method and the importance of hypothetico-deduction in the The Poverty of Historicism:
“I … propose a doctrine of the unity of method; that is to say, the view that all theoretical or generalizing sciences make use of the same method, whether they are natural sciences or social sciences …. I do not intend to assert that there are no differences whatever between the methods of the theoretical sciences of nature and of society …. But I agree with Comte and Mill—and with many others, such as C. Menger—that the methods in the two fields are fundamentally the same (though what I understand by them may not be what they had in mind). The methods always amount to deductive causal explanation, prediction, and testing, as sketched in the foregoing section. This has sometimes been called the hypothetico-deductive method, or more often the method of hypothesis, for it does not achieve absolute certainty for any of the scientific statements which it tests; rather, these statements always retain the character of tentative hypotheses, even though their character of tentativeness may cease to be obvious after they have passed a great number of severe tests” (Popper 1976: 130–131).
I will note in conclusion that a Critical Realist methodology is often proposed for Post Keynesianism (King 2002: 197–200; Jespersen 2009), and that this position might be compatible with the core elements of a Popperian methodology too (King 2002: 253–254; Jespersen 2009: 57–62), and that in his later work Popper appears to have moved closer to Critical Realism (Lawson 1999: 8–9).


I noted above that an intuitive ability to reason inductively is probably an innate trait of the human mind, given to us by evolution by natural selection. Animals, for example, appear to use rudimentary induction in probability calculations when they forage for food (Real 1991).

Even though there is no rational justification for it, why then has induction been so successful and obviously selected for as a survival trait? Even if Popper and Hume are right, we are faced with the paradox that we appear use induction very frequently and that it is a successful form of reasoning, by and large (the fact that induction is of limited use in non-ergodic stochastic systems and that we can err in our inductive arguments does not change this fact [for a list of common fallacies in defective inductive reasoning, see Copi and Cohen 2005: 140–145]).

That inductive reasoning has been highly successful and useful in increasing our chances of survival (and was thus selected by evolution) does not necessarily mean that is it rationally justified, of course.

For example, the belief in life after death might very well help some human beings overcome the trauma and stress of losing a loved one or even contemplating their own death (perhaps it might even make them healthier and better able to survive?), but it is still an irrational, unjustified idea.

But the success of induction is presumably explained by the uniformity of nature which has persisted since the first living things with rudimentary inductive reasoning evolved (for example, if the law of gravity had stopped working 1 million years ago, then all land animals would simply have floated off the planet into space and been killed, and there would be no human beings today; the fact that we are here strongly suggests that basic natural laws have remained uniform).

Human beings have also evolved to use induction. I am aware that this is somewhat similar to Willard Van Orman Quine’s (1908–2000) naturalized epistemology and his explanation of why induction is successful. Quine argues that, while Darwinian evolution does not justify induction, it must explain why it is so effective (Derksen 2000: 27–28; Quine 1975; as an aside, Quine also argues that there is no synthetic/analytic distinction. Thus analytic propositions are just firmly-held synthetic propositions, and there are no real a priori propositions. Quine agrees with Popper that the essence of science is the falsificationist hypothetico-deductive method). And, if the uniformity of nature continues to hold in the future, then presumably inductive reasoning will continue to be successful in those areas where it works well.


Akerlof, G. A. and R. J. Shiller, 2009. Animal Spirits: How Human Psychology Drives the Economy, and Why It Matters for Global Capitalism, Princeton University Press, Princeton.

Barkley Rosser, J. 2001. “Uncertainty and Expectations,” in R. P. F. Holt and S. Pressman (eds), 2001. A New Guide to Post Keynesian Economics, Routledge, London and New York. 52–64.

Blaug, M. 1992. The Methodology of Economics, or, How Economists Explain, Cambridge University Press, Cambridge and New York, NY.

Boland, L. A. 2003. “Dealing with Popper in Economic Methodology,” Philosophy of the Social Sciences 33: 477–498.

Boland, L. A. 2006. “Seven Decades of Economic Methodology: A Popperian Perspective,” in I. Jarvie, K. Milford, and D. Miller (eds), 2006. Karl Popper: A Centenary Assessment. Volume III. Science, Ashgate Publishing Ltd., Aldershot, Hants, England. 219–227.

BonJour, L. 1998. In Defense of Pure Reason, Cambridge University Press, Cambridge.

Caldwell, B. J. 1985. Beyond Positivism: Economic Methodology in the Twentieth Century, G. Allen and Unwin, London and Boston.

Copi, I. M. and C. Cohen. 2005. Introduction to Logic (12th edn), Pearson/Prentice Hall, Upper Saddle River, N.J.

Davidson, P. 2002, Financial Markets, Money, and the Real World, Edward Elgar, Cheltenham, UK.

Davidson, P. 2004. “Uncertainty and Monetary Policy,” in P. Mooslechner, H. Schuberth, M. Sch├╝rz (eds), Economic Policy under Uncertainty: The Role of Truth and Accountability in Policy Advice, Edward Elgar, Cheltenham, UK and Northampton, MA.

De Marchi, N. (ed.). 1988. The Popperian Legacy in Economics: Papers Presented at a Symposium in Amsterdam, December 1985, Cambridge University Press, Cambridge and New York.

Derksen, T. 2000. “Naturalistic Epistemology, Murder and Suicide? But what about the Promises!,” in L. Decock and L. Horsten (eds), Quine: Naturalized Epistemology Perceptual Knowledge and Ontology, Rodopi, Amsterdam, 2000. 15–34.

Dunn, S. P. 2008. The ‘Uncertain’ Foundations of Post Keynesian Economics, Routledge, London.

Edwards, P. 1965. “Bertrand Russell’s Doubts about Induction,” in A. Flew (ed.), Logic and Language, First and Second Series, Doubleday, Garden City, N.Y. 59–85.

Evans, J. St. B. T. 2007. Hypothetical Thinking: Dual Processes in Reasoning and Judgement, Taylor & Francis, Hoboken.

Gerrard, B. 1994. “Animal Spirits,” in P. Arestis and M. Sawyer (eds), The Elgar Companion to Radical Political Economy, Elgar, Aldershot. 15–19.

Glickman, M. 2003. “Uncertainty,” in J. E. King (ed.), The Elgar Companion to Post Keynesian Economics, E. Elgar Pub., Cheltenham, UK and Northhampton, MA. 366–370.

Jespersen, J. 2009. Macroeconomic Methodology: A Post-Keynesian Perspective, Edward Elgar Publishing Limited, Cheltenham.

Lawson, T. 1999. “Developments in Economics as Realist Social Theory,” in S. Fleetwood (ed.), Critical Realism in Economics: Development and Debate, Routledge, London and New York. 3–20.

Keynes, J. M. 2008 [1936]. The General Theory of Employment, Interest, and Money, Atlantic Publishers, New Delhi.

King, J. E. 2002. A History of Post Keynesian Economics since 1936, Edward Elgar Publishing, Cheltenham, UK and Northampton, MA.

Milonakis, D. and B. Fine, 2009 From Political Economy to Economics: Method, the Social and the Historical in the Evolution of Economic Theory, Routledge, New York.

Musgrave, A. E. 2004. “How Popper (might have) Solved the Problem of Induction,” in P. Catton and G. Macdonald (eds), Karl Popper: Critical Appraisals, Routledge, Abingdon, Oxon, England. 16–27.

Parsons, S. D. 2003. “Austrian School of Economics,” in J. E. King (ed.), The Elgar Companion to Post Keynesian Economics, E. Elgar Pub., Cheltenham, UK and Northhampton, MA. 5–10.

Popper, K. 1976. The Poverty of Historicism (1st edn 1957; 2nd corrected edn 1961), Routledge and Kegan Paul, London.

Popper, K. R. 1983. Realism and the Aim of Science, Rowman and Littlefield, Totowa, N.J.

Quine, W. van O. 1975. “The Nature of Natural Knowledge,” in S. Guttenplan (ed.), Mind and Language, Clarendon Press, Oxford. 67–81.

Real, L. A. 1991. “Animal choice behavior and the evolution of cognitive architecture,” Science 253: 980–986.

Runde, J. 1999. “On Popper, Probabilities and Propensities,” in S. Fleetwood (ed.), Critical Realism in Economics: Development and Debate, Routledge, London and New York. 63–82.

Salmon, W. 1974. ‘The Pragmatic Justification of Induction’, in R. Swinburne (ed.), The Justification of Induction, Oxford University Press, Oxford. 85–97.

Skousen, M. 2005. Vienna & Chicago, Friends or Foes?: A Tale of Two Schools of Free-Market Economics, Capital Press/Regnery Pub., Washington, DC.

Strawson, P. F. 1952. Introduction to Logical Theory, Methuen & Co. Ltd, London.

Will, F. L. 1953. “Will the Future Be Like the Past?,” in A. Flew (ed.), Logic and Language: Second Series, Blackwell, Oxford. 32–50.


  1. I hate to be that post-modern guy, but the word "uncertainty" is a euphemism for "disaster is coming but we don't want people to listen to the cranks that were smarter than us".

  2. This is a fantastic article.

    Reading it, I was wondering if you had any thoughts on Nassim Taleb's "Black Swan" or "Fooled by Randomness", as he is interested in a lot of these ideas.

    If you have read any of his books, do you agree what he writes?

  3. I have to admit I have not read Nassim Nicholas Taleb's work.

    But there is a useful article here by the Post Keynesian economist Paul Davidson that discusses it:

    Davidson thinks Taleb’s Black Swan Theory is just a new form of Frank Knight’s concept of uncertainty (see Knight, Risk Uncertainty, and Profit New York, 1921).

    What is crucial is the difference between "epistemological uncertainty" in an ergodic system and "ontological uncertainty" in a nonergodic system. The economy is a nonergodic stochastic system, and we face fundamental uncertainty about many events in the future.

    But to be honest, I really have to read Taleb's work - I will put it in my "to read" list.

  4. Another article worth looking at on Taleb’s Black Swan Theory from the Post Keynesian perspective is:

    Andrea Terzi, 2010, “Keynes’s uncertainty is not about white or black swans,” Journal of Post Keynesian Economics 32.4: 559–566.

    Taleb’s “black swan events” are essentially ones whose probabilities are incalculable.

    But long before Taleb, Frank Knight, J. M. Keynes, George L.S. Shackle were stressing that we face fundamental uncertainty about many events in the future.

    In fact, Taleb’s theory doesn’t go far enough:

    “Keynes’s intractable uncertainty is a deeper matter, and Taleb’s swans seem to float on the surface of it. Keynes used this concept to describe a world where the future outcomes of today’s economic decisions are largely influenced by agents’ current behavior and beliefs. … Paul Davidson has reformulated Keynes’s assumption of uncertainty in terms of the nonergodic hypothesis. Uncertainty cannot be permanently reduced, but the consequences of uncertainty can be controlled by devising institutional arrangements … and using the properties of a sovereign monetary system. System robustness increases when public policies address a system’s failures through the provision of demand and liquidity” (Terzi 2010: 564).

    Taleb’s Darwinian approach is self-defeating in a non-ergodic system with ontological uncertainty.

    “survival of the fittest is a self-defeating approach under ontological uncertainty: in a nonergodic world, a strategy that succeeds under given Keynes’s Uncertainty is Not About White or Black Swans conditions may fail under different conditions. To pursue the analogy, even Taleb’s black swan argument may face its own “black swan”! (pp. 564-565).

    So I would have to say that the Post Keynesian concept of uncertainty is better than Taleb’s.

  5. And, also, the policy prescriptions of Post Keynesian economics to deal with ontological uncertainty are superior.

  6. I will also note that the neo-Austrians are subject to a very similar flaw in their thinking.
    The Austrians believe in epistemological uncertainty, not ontological uncertainty (as in Post Keynesian economics).
    The Austrians think there is some kind of market process analogous to evolution by natural selection that produces pattern/plan co-ordination in the free market (their version of the neoclassical concept of “equilibrium”).
    But in the face of ontological uncertainty and subjective expectations that argument just doesn’t work.

    Some further reading:

    Paul Davidson, 1989. “The Economics of Ignorance or Ignorance of Economics?,” Critical Review 3.3/4: 467–487

    McKenna, E. J. and D. Zannoni, 1997/1998. “Post Keynesian economics and the philosophy of individualism,” Journal of Post Keynesian Economics 20.2 (Winter): 235–249.

  7. Also, a correction to the last quotation of Terzi above:

    "in a nonergodic world, a strategy that succeeds under given conditions may fail under different conditions. To pursue the analogy, even Taleb’s black swan argument may face its own “black swan”!"

    Andrea Terzi, 2010, “Keynes’s uncertainty is not about white or black swans,” Journal of Post Keynesian Economics 32.4: p. 564-5.

    So can there be any real and successful evolutionary process of selection in such a system?

    I think the problem is that the economy is a nonergodic stochastic system.
    Our environment in the past in less complex/primitive human societies was probably not (but maybe I am wrong).

  8. I'll add one more comment, before I write a whole new post here!

    For more on the misuse and abuse of the concept of Darwinian evolution in economics and the misuse of it in attempts to justify the free market, see:

    Stephen P. Dunn, 2008. The “Uncertain’ Foundations of Post Keynesian Economics: Essays in Exploration, Routledge, Abingdon, Oxon, England. pp. 135ff.

    Geoffrey M. Hodgson, 1993. Economics and Evolution: Bringing Life back into Economics, Polity Press, Cambridge.

    Geoffrey M. Hodgson (ed.). 2009. Darwinism and Economics, Edward Elgar, Cheltenham, UK and Northampton, MA.

    Geoffrey M. Hodgson, 1999. Economics and Utopia: Why the Learning Economy is not the End of History, Routledge, London and New York.

    Geoffrey M. Hodgson, 1994. “Optimisation and Evolution: Winter’s critique of Friedman Revisted,” Cambridge Journal of Economics 18: 413–430.

    Basically, there is no reason to think that markets operating in an alleged “evolutionary” manner will lead to economic optimality or efficiency.

  9. Thank you for the comprehensive responses!

    I will add these to my reading list.

  10. Complexity theory is probably relevant to induction. Uncertainty and complexity live side by side. Induction may be unjustifiable in the context of uncertainty and yet be efficient in relation to complexity.