Every society (to the extent that there is such a thing) has its own set of values based on which decisions are made. Let us call the place for those decisions the economic system, and secondly, let us assume that this system follows a dynamic model. It can then be said that the system is ergodic when the set of values underneath have been accurately incorporated into the systemic model and have sufficient explanatory value so as to conduct decisions based on it 'ad infinitum'. In that case, there is ergodicity.
Of course this is an immense oversimplification because it assumes that a system can be explained by a single model, which is certainly not the case at macro-economic level. The logic serves the purpose of this article.
Yet there are also systems that are non-ergodic, in which events take place that are not sustainable ad infinitum. One such example, which we have derived from the book 'Skin in the Game' by Nassim Taleb, is Russian Roulette (Skin in the Game, 2018). The situational description can be found on p. 225. It is said that when a participant in Russian Roulette keeps firing the revolver at its head, at some point he will die from the bullet, even if for a long while he has 'only' conducted a mock execution on himself without being hit with a bullet. The danger in that case would be to conclude early on that Russian Roulette is safe and stable to play over time, because the participant did not receive any bullet yet.
This is an extreme example, but the logic can also be applied to econometric models. Suppose we have the following Ordinary Least Squares model: Y = B0x + B1Y + B2Y + e, in which B0x is a constant, B1Y and B2Y are variables and 'e' is the error term.
In itself, this is a static model, and a very simple one. But then suppose this regression is being repeated every year, and that suddenly something happens in the error term, a sudden bump that makes it immensely large, which violates at least one of the six OLS assumptions and cannot easily be explained away. In that case, it can be said that the model is no longer ergodic, the overall input for the variables is such that repeating a regression as per this model over time, leads to less and less accurate estimates of Y and perhaps in the long-run even to a situation, in which Y can no longer be meaningfully explained.
Yet what are the causes for such bumps? One could for example think of a whistleblower, or other individual who suddenly discovers billions of assets that have been laundered or siphoned off the economic balance sheets of systemically important banks. And suppose that this person has fled the system, under threats and censorship, so that his or her findings could not have been included in the variables in any meaningful way? In that case there are two problems that could arise besides serious distress to the person involved: i) large standard errors, significantly reducing overall explanatory power of the variables ii) multicollinearity or some form of serial correlation because a major information input is ignored that relates to multiple variables in the model. Worst case there is genuine omitted variable bias.
We can derive two conclusions from this: i) The error term can become so large that the explanatory value of the model vanishes completely, leaving us in the dark and unable to use its input to make economic decisions in such a way that the system can be sustained ii) There is an important relationship between whistleblowers and economic ergodicity. Taking this into account not only fosters democracy and human rights, but also helps to guard institutional stability, a major precondition for a functioning economy.
Equally important is the observation that non-ergodicity leads to a conflict with natural law/eternal law and thus with liberalism. Hence ergodicity is required to align a system with natural law and uphold a free society.
Add comment
Comments