Dear Pedro, Rob, and others,

As I read the Treatise on Probability it appears to me at its heart to be an attempt to outline the approach to inference that later developed into “Bayesian” probability using language that Keynes thought might be acceptable to the philosophical tradition of the Cambridge of his time (and, given lags and the Great War, perhaps ten or fifteen years before the publication of the Treatise). The important themes of the Bayesian approach are a “subjective” view of probability as a statement of the degree of belief of a decision maker given particular evidence (or information), and the importance of “prior” judgments as to possible hypotheses in disciplining quantitative inference. Keynes characteristically emphasized the many factors that make the application of this model to social and economic phenomena, particularly inferences and predictions about macroeconomic fluctuations and financial asset prices (which Keynes had been observing as a participant for some time) problematic.

The two central elements in Keynes’ critique seem to me to be the difficulty in establishing a coherent prior over social and historical events, and the ensuing incoherence of trying to express inferences and judgments in quantitative form as numerical probabilities. I don’t think Keynes had much quarrel with the Bayesian (or Laplacian) analysis of games of chance or scientific data generated by observation of nature, but had severe doubts about the applicability of these methods to human historical sciences, of which economics (or political economy) is an example.

A particularly relevant issue in this respect is the interpersonal aspect of prior formation and the process by which communities form consensus on priors, which Keynes seems to have thought of as “conventions”.

Keynes’ views on these fundamental issues do not seem to have changed much over his career. He consistently deprecated the idea of what we now think of as “econometric” estimation of probabilities over complex social and economic events as a kind of fantasy. Keynes was intensely interested in economic data and a pioneer in the organization of national income accounts so his attitudes were far from anti-empirical. His philosophical investigation of probability seems to have convinced, him, however, that the project of applying systematic theories of quantitative inference at aggregate levels was a will o’ the wisp.

Duncan Foley

> On Aug 12, 2015, at 5:31 AM, Rob Tye <[log in to unmask]> wrote:
> 
> Dear Pedro,
> 
> OK – I take your point.  But the sort of suggestion Wittgenstein seems to be
> making in that LP quote, and later much expanded on, seems to have been one
> just as clearly in the mind of the early Keynes himself, so to me, the
> question reads rather like: ‘did the early Keynes influence the later
> Keynes’ (answer: yes).  Skidelsky mentions a close intellectual relationship
> between the two began on 30th October 1912.  Wittgenstein was then new to
> philosophy, but Keynes had by then been working on exactly these matters for
> 6 years.
> 
> The Walley work certainly builds on a position taken in the later part of
> the Keynes Treatise.  But it, (like Keynes himself) seems to contradict what
> is very prominently stated in the first part of the Treatise.  If Keynes
> meant ‘fuzzily numerical’ by ‘not numerical’, he surely owed it to his
> reader to say that clearly from the onset?
> 
> Best Wishes
> 
> Rob Tye

Duncan K. Foley
Leo Model Professor
New School for Social Research
Department of Economics
6 East 16th Street, D-1127
New York, NY 10003
(212)-229-5717 x 3177
fax: (212)-229-5724
e-mail: [log in to unmask]
alternate: [log in to unmask]
webpage: https://sites.google.com/a/newschool.edu/duncan-foley-homepage/home