plumbing is not architecture

systems, change and the delusions of agency

Plumbing is important, don’t get me wrong; in a sense it was plumbing that caused the fall of the Roman Empire.  And in a financial crisis (which is the context in which I’m usually tearing my hair out and telling people “don’t mistake plumbing for architecture”), the “financial plumbing”, in the sense of the legal and institutional arrangements of money markets, can matter a lot – it can render someone unable to make a payment that they might otherwise have been able to make, and this can lead to a horrifying cascade of legal and institutional consequences.

But if you ever walk round an old building with your eye out for such things, you’ll often see bits of pipe sticking out the wall which seem to go nowhere.  That’s a sign that a pipe has been “capped off”.  At some point in the past, somebody rerouted things.  And that’s why I think the term “financial plumbing” is one of those metaphors and jokes that you can gain insight by taking literally.  The thing about plumbing is that it can be changed.

This fact about plumbing reliably infuriates people who learn a lot about it in the belief that it will help them understand global macro.  The trouble is that in normal times, the plumbing “just works”, and doesn’t really interfere with anything else.  In big crises, the rules get rewritten and things which might have been considered impossible from a financial plumbing point of view become not just possible, but normal.  There’s only a short transitional period during which knowing about plumbing is useful – when the constraints are just beginning to bite, but before they become such a big problem that it becomes someone’s priority to do something about them.

I think this concept generalises to a lot of other plumbing-like systems, which seem like they’re part of the essential structure, but which are actually utilities hanging off the real structure and subject to change and cancellation at short notice.  For example, if you understand a lot about bank regulation, it might be very reasonable to assume that because bank capital requirements make a big difference to lending standards, they could be used as a tool of environmental policy.  Or if you understand a lot about payment systems, you might think that because everyone uses them all the time, they would be a devastating weapon of economic war.  You might think that access to the British legal system was essential for repressive foreign governments.

And you’d be repeatedly disappointed.  Every time you tried to use one of these things as a lever to do something big, it would suddenly turn out to be much less effective than it seemed to be when you were using it to achieve more modest goals.  The key problem would be something analogous to Goodhart’s Law – the system has the ability to change itself, and the way in which it changes itself will often be such as to accommodate the things you were trying to do.

What I’m talking about here is a property that was mega-important to the early writers on cybernetics – the Macy Conference types, Ross Ashby and the like.  Ashby called it “ultra-stability”; the ability of a system to reorganise the relationships between its parts in order to re-establish equilibrium when it is broken in the short term.

Ultra-stable systems usually achieve that property by having multiple levels of control; a fast-moving system that responds to immediate environmental stimuli, and a slower-moving regulatory system that changes the parameters of the fast system.  When the fast system goes out of equilibrium, it triggers the slow system, which then tries different things until it reaches a state it can live with (or alternatively, goes on cycling forever or blows up – ultrastable doesn’t mean invulnerable!).

Most real world viable systems are ultrastable in important ways.  They can reorganise their plumbing in order to continue to serve their architecture.  You need to understand the plumbing, but also to understand that it is plumbing, and to have at least some idea of how the higher-level system will react.  In general, trying to change a big system by playing around with its plumbing is a mug’s game.