Five systems laws and what they tell us about the financial crash: Part 1


Five systems laws and what they tell us about the financial crash: Part 1

In 2008, Her Majesty the Queen asked academics at LSE why nobody had seen the financial crash coming. It was four years before she got anything resembling an answer from an economist and then the answer seemed to boil down to observing that everyone seemed to think it was someone else’s job to look out for it.

If instead of taking a classical economics view of the problem, we look at it from a systems thinking perspective, we get some somewhat different answers, but significantly we get them rather faster than having to wait four years. I did this exercise with a group of MBA students a couple of years after the crash. They were new to systems theory, but not to economics and some were themselves from the banking and financial industries. One of the unusual things about systems approaches is that they are underpinned by a set of laws and theorems – many of which have mathematical proofs. Using a set of these laws the students were able to generate a set of mutually consistent and coherent explanations in less than an hour. In these two blogs, I take five of those systems laws and apply them to the financial crash to see what they can tell us about both why it happened and whether it was indeed predictable. This is in no way exhaustive – there are other systems laws we could have picked and those might have yielded other answers, so this doesn’t purport to be a complete use of systems approaches to the problem, nor a complete answer to it, but it does show the sort of alternative views that can be produced – and very quickly – from applying systems laws to complex problems.

The five laws I’ve picked are:

  1. Feedback Dominance Theorem: For high gain amplifiers, the feedback dominates the output over wide variations in input
  2. Conant Ashby Theorem: Every good regulator of a system must be a model of that system
  3. Redundancy of Potential Command Principle: In any complex decision network, the potential to act effectively is conferred by an adequate concatenation of information
  4. Adams 3rd Law: A system composed of the lowest risk components available will be a high risk system
  5. Relaxation Time Principle: System stability is possible only if the system’s relaxation time is shorter than the mean time between disturbances

So let’s take the impact of each of those in turn.

  1. Feedback dominance theorem: For high gain amplifiers, the feedback dominates the output over wide variations in input

Our economy is built around what in systems terms is called a “high gain positive feedback amplifier”. In this context, “Positive feedback” doesn’t mean telling people you like them, it means that an output from part of the system feeds-back to increase the system that generated the output. It’s the technical description of a system that feeds on itself in a “virtuous circle” or a “vicious circle”. In this case, money deposited with a bank gets lent out to borrowers who invest or spend it and the people they spend it with put it into … a bank. The “high gain amplifier” bit is that an initial deposit of £100 can through the process of lending and relending turn into £1000. In fact it’s even more amplified than this because it doesn’t even require an initial deposit, the process can (and does) start with the bank lending and then seeking deposits to underpin its lending.

In systems, positive feedback systems run with ever increasing levels of throughput until they hit some sort of constraint – either there is some regulatory process that can turn down the level of amplification, or they run out of control and crash. A crash generally happens when the system runs out of supply. In this case, the economy first ran out of safe borrowers to lend to, so to keep it growing, loans were offered to unsafe borrowers – creating the sub-prime market. What finally ran out and precipitated the crash was a combination of confidence and the overall liquidity in the system – just too much had been lent out relative to the “real” cash in the system.

The scariest part of the feedback dominance theorem is the bit about the “feedback dominates the output over wide variations in input”. What this means is that if you have the same feedback structure, then you will always get to the same endpoint no matter where you start. Given the feedback structure, the crash was pretty much inevitable.

After the crash, governments and regulators worked hard to rebuild the feedback structure and restock it with fresh money so it could start up the process again. In other words, they were working to rebuild an economic structure that had just crashed and that is designed in such a way that it must crash again. In fact of course there have been repeated crashes in economic history, what we experienced wasn’t remotely unusual. The reason the powers that be set about getting the economy back on its course is that until it crashes, the feedback generates growth and that makes people prosper and actually economists don’t know how to run an economy in any other way other than continuous growth. Crashes are uncomfortable and have significant impact on people’s lives, but the harsher reality is that it is impossible to run any system with unlimited growth indefinitely in a finite environment and the world economy is starting to hit the limits of supply in several key areas. IF we try to run an unlimited growth economy to the point where it doesn’t just run out of confidence and liquidity, but instead runs out of basic resources, there be no way to restart that.

  1. Conant Ashby Theorem: Every good regulator of a system must be a model of that system

Conant Ashby’s theorem, also commonly referred to as the “good regulator” theorem, means that you can’t manage what you can’t understand and your ability to understand how a system works depends on how good the models you have of it are. Simple, even obvious, but profound and frequently ignored. So, in the crash, the regulators didn’t have a model of how all the new financial instruments worked and particularly, they didn’t have a model of how they were interdependent. As a result, when the liquidity problems stared to bite and people started to lose confidence, the chain reaction of collapse was unexpected. It had not been understood how much had been risked and how those risks had been piled one on another. What was and was not a risk was invisible to regulators as risks were disguised as assets and traded as such. Essentially, large areas of the global economic system were invisible to regulators – they were not accurately represented in the regulators models of the economy. In terms of Conant Ashby then this was an unregulated system, one in which nobody really knew what was going on. At some stage, someone had understood how each of the elements in the system worked, but nobody understood how the whole system worked.

In the UK at least, the financial regulators (including the treasury) have always relied on non-dynamic linear models to model a dynamic system with feedback loops – this can never be “Conant Ashby compliant”.

We’ll pick up Laws 3-5 in a subsequent post.