While the Heterodox Economics Newsletter regularly collects recent...
While the Heterodox Economics Newsletter regularly collects recent issues of journals dedicated to heterodox economics to some degree, it is often the case that contributions of potential interest to heterodox economists are published outside this, necessarily selective, spectrum of journals. I came across one such case in a recent issue of Nature Physics, which featured a paper on ergodicity and its role for economics (see also the related editorial). Not being a specialist in this, my intuitive understanding of (non-)ergodicity always was that, if processes are non-ergodic, then estimating some expectation value (e.g. a mean) by considering the probability of all possible states (‘ensemble average’) will give different results as compared to evaluating states over time (‘time average’). Hence, as the argument goes, as soon as processes are considered non-ergodic, time has to be explicitly considered in both, reasoning and modeling. In this vein it is sometimes added that ‘real time’ and ‘history’ matter insofar as future outcomes will depend on past results, which may give rise to path-dependency, non-linearity and fundamental uncertainty (as we can’t judge the future from the past).
One very simplified application of this argument is to say that some economic decisions (e.g. on investments) are different from the lotteries used in expected utility, as the former are made only once and, hence, the implicit ‘pooling’ of decisions inherent in expected utility theory is invalid (as, among others, argued by Shackle). By doing so we introduce ‘time’ in a very basic and specific way by emphasizing the irreversibility of decisions, which is supposed to render the expected value irrelevant. What I learned from the Nature -paper is now, that (1) non-ergodicity does not necessarily lead to uncertainty and (2) that it is actually easy to illustrate the basic intuition that time matters in a simple way without emphasizing the uniqueness of events. For this purpose the paper uses an example of stochastic process that governs the accumulation of wealth. In this example the rate of return follows a (slightly biased) random walk, which can either take a positive or a negative value. This process is non-ergodic and the author, Ole Peters, shows that an explicit dynamic consideration of the problem gives an accurate result, while expected utility theory will not (as it will take the average of all possible outcomes in any single round of the gamble instead of considering the evolution of wealth explicitly).
Retrospectively, I think this finding is rather intuitive (as are other examples on the related research blog, which I would also recommend) and some may say it’s obvious anyway. In any case I think it is a great idea for illustrating the principle underlying the problem at stake - namely that time matters.
Hope this helps and all the best,
Jakob
PS: As often in such cases of potential interdisciplinary cross-fertilization caution and patience play a key role. I would, for instance, assume that many heterodox economists are skeptical with regard to the notion of optimization found in the paper. At the same, the author would probably be skeptical about the link between non-ergodicity and uncertainty mentioned above. Nonetheless, there is a common anchor – time matters – and it’s still an open question what exactly to make of this common anchor.