I posted a paper by Carlos Zappia on how financial economics is likely to be more Keynesian kind post crisis. His take was much of financial economics and decision theory is based on Bayesian probability. Bayesian in turn does not really model uncertainty which is core problem in financial decisions. What is instead needed is a behavioral decision theory which looks at the idea of uncertainty.
Work has started already on modeling these new paradigms. However, none of this is new as Keynes meant the same in his animal spirits ideas. So it is going to resemble Keynesian ideas.
In another paper he expands the idea further.
A by-product of the recent financial crisis has been the renewed interest in Keynes’s works. Both in the press and in scientific journals, a crowd of commentators has emphasised the need to scrutinise the General Theory in order to gain a better understanding of the actual macro-dynamics of the economy and of the policy measures apt to help the economy recover from the downturn. But Keynes’s thought has been given central prominence also with respect to the understanding of what went wrong at the microeconomic level, with specific reference to the role played by “irrational” agents animated by animal spirits.
This paper supplements the influential analysis of George Akerlof and Robert Shiller’s Animal Spirits by arguing that a Keynesian explanation of the actual behaviour of individual agents is to be based more on the Treatise on Probability than on the General Theory itself. Indeed, while it is well-know that the rationale of Keynes’s rejection of “Benthamite calculus” is best provided in the Treatise, less attention is usually given to the constructive analysis emerging from his criticism of contemporary probability theory.
Through an assessment of Keynes’s examination of “the application of probability to conduct” in the Treatise, the paper shows that most of the developments of what is usually referred to as behavioural finance have a Keynesian origin. In particular Keynes hinted at a decision rule different from mathematical expectation, a rule intended to mimic the behaviour of actual agents making decisions under uncertainty. The understanding of the current financial crisis, the paper concludes, would gain from a Keynesian assessment of the rationale for actual decisions as much as from the usual one concerning macroeconomic policy.
My know-how on the subject is limited hence no comments. Though, have maintained Keynes contribution to eco is far more than government stimulus.
This paper is clearly in this spirit helping us understand his contribution to probability.
The author summarises the lessons as:
In the above discussion it has been detailed that a fresh reading of the Treatise can help clarify a number of issues raised by commentators of the actual crisis when they discuss the role of individual agents. First, Keynes’s understanding of “non-numerical” probabilities hints at the notion of decision weights associated in current economic theory with Kahneman and Tversky’s (1979) rationalisation of actual behaviour through weights that are probability measures that do not satisfy the property of additivity (Tverski and Kahneman 1992). Second, the degree of non-additivity of the probability measure introduced by Schmeidler (1989) can account for confidence in a probability assessment without any reference to a second order probability, as suggested by Keynes’s weight of argument. Third, certain criteria for decision making under uncertainty put forward in the non-additive literature are devised to incorporate a measure of the degree of confidence in the probability assessment, also an issue Keynes discussed in the Treatise.
The third point is discussed so many times but hardly implemented. Forecasts are given without specifying how confident the person is on the forecast. So he may be saying he expects x to rise by 5% but if he is just 50% confident, you know the forecast is pretty uncertain.
We are clearly living in multimodal times as highlighted by PIMCO guys so often. Multiple adverse situations are likely to happen with relatively similar probabilities. One needs to be careful with assigning blank probabilities in these times.