Reader Richard Kline, who provides regular, sophisticated comments, was keen to continue the discussion provoked by a post last week, “Hoisted From Comments: Greater Liquidity Produces Instability.” An anonymous reader offered a complex systems theory view of our modern financial system. The opening paragraphs:
Perhaps a lesson to be learned here is that liquidity acts as an efficient conductor of risk. It doesn’t make risk go away, but moves it more quickly from one investment sector to another.
From a complex systems theory standpoint, this is exactly what you would do if you wanted to take a stable system and destabilize it.
One of the things that helps to enable non-linear behavior in a complex system is promiscuity of information (i.e., feedback loops but in a more generalized sense) across a wide scope of the system.
One way you can attempt to stabilize a complex system through suppressing its non-linear behavior is to divide it up into little boxes and use them to compartmentalize information so signals cannot easily propagate quickly across the entire system.
I hope I am not oversimplifying what either the anonymous reader or Richard intend to convey, but the non-linear issue is not trivial. Processes that are described by non-linear equations are unpredictable. That is why, per above, inducing or enabling non-linear behavior is Not A Good Idea.
Worse, non-linear math is really hard, so while lots of mere mortals can model linear processes, it takes high powered skills to deal in non-linear modeling. And you therefore get a second problem: due to computational convenience, most practitioners will try to describe a system using linear models, and if it works well enough in most cases, it gets a go. To illustrate: pretty much every mainstream financial model (Black Scholes, for instance) assumes continuous markets, which simplifies the math. This, for instance, is the origin of the classic fat tails problem. Pretty much everyone knows that models that use a normal distribution underestimate tail risk (the odds of outliers, which in this case is dramatic price rises or falls). Yet the flawed models are still consulted out of convenience (note I am not saying other models aren’t used, but the reliance on models known to have fundamental shortcomings is considerable).
Richard has provided a through, thoughtful exploration out of some of the issues. After a general discussion, he sets forth five questions and works through the first one. on innovation (note the discussion ranges far beyond the financial markets). Recall that one of the defenses of our current financial mess is that the products were innovative and hasty regulation will curtain other useful advances (this argument is that the products weren’t the problem, it was the practitioners, or in popular terms, “guns don’t kill people, people kill people”). But as Richard illustrates, that level of discussion is simplistic; there are ways to parse the problem that can lead to better thinking about possible remedies.
His ideas on issues 2-5 will come in later posts in this series.
Your comments very much appreciated. I’ve edited his piece slightly to make it a bit less formal.
Now to Richard Kline:
To what extent have nonlinear processes promoted the Securitization Bubble, precipitated its collapse, or prolonged the resulting instabilities in the financial system? I’ll keep the discussion non-technical, i.e. non-mathematical. While I have an informed opinion, I don’t pretend to expertise, and hope to elicit further comment and debate.
While there is evidence for most of my contentions, it isn’t conclusive;. I raise ideas more than offer conclusions. Some general, but valuable, further reading is suggested for those interested. Comment by those with technical background in nonlinear complex systems, especially economic systems, is welcome—but I’m not holding my breath. Though nonlinear dynamics in financial markets received no little initial research ten years ago and more, many of the specialists involved have since been hired into the hedge fund industry where their work has presumably become proprietary. Not only do we not know what they are doing, we don’t even know what they know now; there has been little recent publication of consequence.
To delve into this issue, then, let us first briefly consider financial markets as systemic phenomena. Given their inherent complexity and diversity of inputs, modern financial markets are inherently complex systems with numerous nonlinear phenomena embedded within their actions, that is phenomena whose transformations are not smooth, not continuous, or both. Such overlapping dynamical phase spaces appear less complex than they are because salient stable equilibria within them are defined by firm, cohesive, and above all observable parameters such as priced units of exchange, transaction terms, regulatory limits, and the like. Such firm parameters do typically though not invariably have the virtue of precluding overtly chaotic behaviors in their respective financial event-spaces, and to a degree in the larger interaction systems which contain them. Indeed, while complex systems will often self organize with emergent properties developing within them in consequence, the intervention of human participants in these markets tends to limit or swiftly capture observable systemic properties—or at least that is the idea.
Since these defined and manipulated parameters are of lower dimension than the market processes to which they map, they give the illusion to the observer that markets themselves are more solid and of lower dimension than is really the case, like skin on hot chocolate. This illusion is compounded by the fact that the very large volume of quantitative data regarding finance and markets, including trend analyses beloved by academically trained economists, are presented in linear analytic terms; ants crawling on that skin, if you will. Such linear models tell something regarding ‘what dwells below,’ but less then we often lead ourselves to believe.
Bear in mind, though, that such linear models only map to the nonlinear trajectories and higher dimensions of the underlying event-spaces, if with fair reliability, rather than fully describe them. These are fuzzy, noisy spaces in that they largely describe human behavior which is intrinsically inexact, information which can be imperfect and/or corrupt(ed), and rule-parameters which are not always followed and which do not capture all relevant processes. Phase spaces and their properties are best described as geometric structures with a time dimension which describe relationships whereas our analyses in a modern educated context are overdefined by linear mathematical methods which abstract fixed values. The present conceptual mismatch of methods to phenomena further leads to an insufficient cognitive engagement with systemic and nonlinear processes on their own terms, in economic behavior and elsewhere:
Our tools are yet poorly matched to the natural phenomena we wish to understand. I will pose it as a truism that processes which appear disjointed or broadly nonlinear do so when they are viewed from perspectives which are or lower dimensionality than are the structures observed; Flatland views of Squareland trajectories. Tensor analysis may prove sufficient to effectively analyze some complex processes; perhaps. Since most of us cannot execute it competently, nor are the guidelines clear by which to operationalize available data into tensor matrices, we will have to sharpen our ‘complex reasoning’ to make heuristic judgments better suited to the data-events instead. This exercise is valuable in and of itself. It is even more true in considering complex systems than otherwise that as you define your questions you describe the parameter space of your possible answers. So, let’s build some better questions.
From that position, here are five questions recently and variously posed which I find personally interesting:
Does innovation require untrammeled information flow across social/ economic event spaces?
Is the crisis in securitized debt the result of a ‘black swan’ event?
Was the creation of the Securitization Bubble the result of nonlinear processes in the financial markets?
Is a financial event-space optimized for propagation desirable?
If not, what structure or process parameters might improve process outcomes?
Innovation: Does innovation require untrammeled information flow notwithstanding any potential costs to an economy or society of undampened interactive trajectories? Not . . . quite.
The stated assumption that innovation requires untrammeled flows of any kind embeds two misconceptions. First, there is an implied confusion between discovery and innovation. Discovery is just that, finding something not previously understood to exist. Exposure to large bodies of information may raise the probabilities of discovery, but so may improved observation of putatively well-known information. Either way, discoveries are comparatively rare; significant discoveries rarer still.
Second and more fundamentally, the stated assumption conflates innovative design and innovation diffusion. The popular belief is widespread that innovative design results from ‘throwing many ideas up against the wall and seeing what sticks;’ that no one really knows what they are doing so innovative ideas and designs are both essentially fortuitous and random. And certainly fortuitous and random innovations do occur. What is required, then, from this perspective is the largest possible supply of things to throw up against the wall. In fact, much the opposite is the case. To cite Edison’s well-known dicta as a benchmark, “Genius is 1% inspiration and 99% perspiration.” This overstates the case, but innovative design tends to happen in small environments which can be effectively modeled to the point where changes from shifts in composite parameters can be approximated hypothetically, additional variables or inputs can be added to the context in a controlled fashion, or both.
Engagement with those environments—i.e. knowledge and skill—tend to improve the frequency and coherence of designs, to which quality of outcome correlates. Fortuitous manipulations do happen, yes; information putatively extraneous to context can provide valuable guidance or comparison, again yes. Innovative design does not necessarily flourish in noisy environments maximally in flux. There, relationships can be hard to grasp, and innovations may soon be suboptimal in ever changing contexts; indeed, conservative but stable designs may better reward success. In brief, innovative design occurs best in the enriched niche, not in the middle of a crossroads.
Innovation diffusion, by contrast, occurs best where information and adaptation are minimally constrained across a context. Consider the adoption of mobile phones in Europe or Korea, where a single technical standard was publicly designated, adoption of mobiles was rapid and deep, and use-driven development burgeoned. In contrast in the US, competing technical
standards and incompatible service provider networks slowed adoption, and have left services fragmented. Diffusion is a process which implies point autonomous use of what is adopted or put to use. In contrast, propagation is more nearly a spread whose nodes remained linked.
Consider Linux an example of diffusion and Windows an example of propagation. Linux point sources can transform or adapt independently, while Windows point sources are under heavy systemic pressures (incompatibility drift) to transform in relation to nodal (i.e. Redmond) based changes. It isn’t commonly understood that many innovative designs are effected well before they diffuse (or are propagated), perhaps because salient fads can diffuse with great rapidity in modern societies. A typical example is the Internet, which was functionally extant well before software refinements turned it into a mass medium, a medium whose greater scales drove product and organizational developments thereafter. The adoption trajectories of innovations most typically are logistic functions in form, but with longer low adoption under-the-radar initial tails then considered, even much longer. Whether relatively rapid diffusion is a social virtue is debatable, but it is certainly an economic gain if only for implementation investment.
There are two points to making this distinction between innovative design and innovation diffusion (or propagation). First, the two processes can be facilitated or inhibited separately. For example, a society with low barriers to diffusion may still be the one to capitalize upon innovations, regardless of source, because they scale the markets and formalize product parameters. Second, large profits come to those investing in innovations which diffuse due to market scale-ups, while huge profits come to those investing in innovations which propagate since they remain substantially intermediated in subsequent capital flows. The arguments one typically hears for lowering barriers to innovation diffusion and damn the consequences are from those hoping that their innovations or the industries tied to them will get the market scale-up opportunities. ‘Pro-adoptionists,’ to give them a name, typically have a stake in the outcome so their perspective is not disinterested (presuming that anyone else’s could be, either). To get innovative designs we need enriched niches whether or not we have low barriers to innovation adoption. We can have rapid adoption without being particularly innovative. Societies can, in fact, deliberately choose whether or not to have rapid adoption.
Moreover and more importantly societies can deliberately choose which innovations to rapidly adopt (within limits); consider China in the latter regard of selective adoption. Choices about which innovations to permit rapid adoption are choices about who will get very rich, however. Much of the shouting about innovation is, at its base, concerned with the last proposition.
Nebojsa Nakicenovic and Arnulf Grübler. 1991. Diffusion of Technologies and Social Behavior.
Jacob Getzels and Mihalyi Csikszentmihalyi. 1976. The Creative Vision.
[Respectively the best texts on technological innovation and the creative process I have ever
found. Of course they are the least known.]