Stability Through Instability

A friend pointed me to a doubly prescient talk given by George Soros in 1994 about his theory of reflexivity in the markets.  Essentially Soros notes that there’s feedback in terms of what agents believe about the market and how the market behaves.  Not groundbreaking, but he takes this thinking to some logical conclusions which are in contrast to standard economic theory:

Instead of a tendency towards some kind of theoretical equilibrium, the participants’ views and the actual state of affairs enter into a process of dynamic disequilibrium which may be mutually self-reinforcing at first, moving both thinking and reality in a certain direction, but is bound to become unsustainable in the long run and engender a move in the opposite direction. The net result is that neither the participants’ views nor the actual state of affairs returns to the condition from which it started.

[W]e can observe three very different conditions in history: the “normal,” in which the participants’ views and the actual state of affairs tend to converge; and two far-from- equilibrium conditions, one of apparent changelessness, in which thinking and reality are very far apart and show no tendency to converge, and one of revolutionary change in which the actual situation is so novel and unexpected and changing so rapidly that the participants’ views cannot keep up with it.

We’ve been discussing the prospect of stabilizing dynamics by intervening during times of “apparent changelessness” so that we can forestall or mitigate times of “revolutionary change”.  Interestingly, there’s (perhaps) no way to tell the difference between the quiet, equilibrium condition and the quiet, far-from-equilibrium condition, but empirically it seems that the former ultimately gives way to the latter — and ultimately to a revolutionary period — it’s just a matter of how long it takes.  As Taleb observes, the longer we go without a black swan event, the more likely its appearance.

Perhaps the difficulty with stabilizing complex adaptive systems has to do with reflexivity.  As soon as we make an explicit policy decision to address a source of instability we know about, the system believes* that it has become more stable, which blinds it to the inherent inevitable falsehood of that proposition.  Which in turn quickens the instability.   Sort of a probabilistic, temporally diffuse liar’s paradox.

This would suggest that any intervention which increases the stability/certainty of the system’s internal representation of itself — i.e. the beliefs of the market participants about the market — actually has the opposite effect as its intent.  Instead, it would be a better approach to induce uncertainty whenever the system seems to be settling into a “quiet” period.  This could be accomplished either by gratuitously creating a limited amount of market volatility, or by obfuscating market-related data.  Given the increasing difficulty with the latter due to technology (not to mention the fairness issues it entails), the former seems preferable.  What would this look like?  It could take many forms, including ones that appear in the comments here.

Ultimately though, what Soros’ arguments suggest to me is that the goal of policy-induced stability is paradoxically better achieved by inducing instability than by attempting to dampen oscillations ala Sumner.

We’ve used the ecological analogy in the past of a controlled burn policy.  Now a biological analogy comes to mind.  If you want to become really strong and resilient, the best way to work out is to put your body into slightly unfamiliar situations and don’t fall into a routine that your body gets used to.

To expect policy makers to do this willingly with the economic policy seems a bit far fetched.  However, interestingly, the historical trajectory of more frequent “revolutionary” periods in the economy may actually have the same effect organically.  That is when individual economic agents (most notably humans) get used to uncertainty being the norm (as opposed to it being a distant or non-existent memory), perhaps the overall economic system will converge to riding that famous “edge of chaos” instead of oscillating in and out of it.

* In speaking of multi-agent systems like markets, when I say it believes something, you can either take that as shorthand for “the participants believe” or you can ascribe cultural agency to the markets as I do.  For the purposes of this discussion, it doesn’t really matter.

  • Related good essay on the Montana Paradox: “the more we try to control and regulate our risk, the more exposed and at risk we are, because the more protected from hazards we think we are, the less conscious of potential dangers we become. What we typically would think of as the safe way to go, which is to provide a lot of control, actually works in the reverse. When we put a lot of controls in place, we rely on those controls and we forget about the notion of risk.”

    Seems reminiscent of Dan Ariely’s TED Talk.

  • I would look at Soro’s actual example of reflexivity, and not his generalizations about it.

    His original example had to do with how mortgage funds in the 70’s were evaluated by the crowd.

    He has an excellent observation about how it is possible to sell a bond with a fixed interest rate for an increasing amount of money.

    The example is well worth reading.

  • John Miller commented to me about the concept of “stability through instability” in a private email. Here’s what he said (posted with permission):

    I think this is a very important point. Given intuitions and even business practice (especially in manufacturing), we feel that “noise” is a bad thing and needs to be removed. Notwithstanding this, there are a lot of situations where noise provides a lot of benefit. For example, for biological evolution you need some variation (truth be told, the vast majority of variation is neutral or harmful, but on occasion good things arise). In optimization problems, you often need to introduce some noise to keep things from getting stuck at inferior optima. In economics, the presence of noise traders will often allow a market to work much better (imagine if everyone relied upon the identical trading rule or expectations—you would get wild price swings). In bee hives, having some noise in the trigger point where individual bees alter their behavior tied to the temperature of the hive, allows much better thermo regulation. I’m not sure how much work has been done on noise and government policy, but I would think similar insights apply (along with a likely mistaken intuition that policy should always be predictable and transparent—probably a good idea in some situations and a bad one in others—I suspect you are right that the dividing line here is closely tied to the complexity of the underlying system).