In early 2008 I was the oldest intern (35) in Westminster, toiling at the Social Market Foundation.  As an experienced financial spiv and prolific economic bluffer, I was asked to present to a roomful of wonks on this thing, “the credit crunch”.*

I wish I could find the slides.  But somehow I had the time to look up Northern Rock’s lending book, and Granite, their securitization vehicle with its countless mortgages. I also dived into indices full of Mortgage Backed Securities, grabbed the codes of securitizations, looked them up at the SEC, and tried to drill down, and then …

And then the complexity defeated me – which was the point.  There was no way I could demonstrate to this roomful of geeks what was actually going on.  When financial markets are brimming all you need to know about the instrument you trade is that it has a few strong bids and offers on a tight spread.  When they are not, you need to understand what you hold, because you may be holding it for a while. But given the endogeneity of the variables (those mortgages would turn out to be strongly influenced by the macro economy) and the role of human agents throughout, reacting to a new situation, such knowledge of the present would be impossible. Don’t even ask about the future.

Around that time the Large Hadron Collider had fired up, spitting out terabytes of data from trillions of subatomic collisions.  I remember remarking that the scientists had it easy; they didn’t have to include in their calculations what each and every particle thought of it all, and how its opinions changed its actions.

It is no coincidence that some of the most notorious SpAds have some of the best written blogs: McBride’s is sadly gone, but was astonishing at times, and now we have Dom Cummings’.  His latest demonstrates swiftly how complexity can build up to  impossible levels from as simple a scenario as a chess game.  Earlier, he tweeted at me an explanation for why economics has (apparently) fallen in esteem – because, to paraphrase atrociously,  it lacks mathematical or empirical rigour.

The tweet is gone, but he later tweeted Von Neumann’s introduction to his 1944 book on game theory, which is excerpted here (around page 23) in Cummings’ long piece. It has this splendid quote:

“An almost exact theory of a gas, containing about 10^25 freely moving particles, is incomparably easier than that of the solar system, made up of 9 major bodies”

Imagine a rectangular water-filled bathtub full of bobbing apples.  If you are asked to predict exactly where each apple will be in five minutes, you will fail. But if you are asked to predict what the water level will be if you poured in another litre , all you need is its dimensions and you can get it right to within a millimetre.  Extremely complex systems can be understood in certain limited ways.  Exact micro forecasts are not the test of the scientific nature of a discipline, and nor are ever more complex models.  Macro economics should surely not strive to emulate the Meteorological office, where predicting down to a much lower level is what is needed.

Cummings’ post is about far more than this, and so the point I make is not a step in any specific argument.  But in his conclusion he stretches the argument rather far when he implies that a lack of appreciation of complexity theory is behind the over-confidence of a Balls or a Cameron.  Their overconfidence must be as much about survivorship bias – the chutzpah of people who have been promoted at a daunting speed to positions of unbelievable authority.

His point about economics is a popular one however.  How can the discipline retain credibility when complexity frequently defeats any attempts at forecasts? When the likes of Greenspan were so confident before, and complexity showed up the hubris, haven’t we learned something rather damming about economics?

I take the unfashionable view. Yes, a decade of smooth growth led to confidence, but it wasn’t the confidence of people who thought they’d conquered complexity. It reflected a belief that policymakers could and would control the broad macro economic conditions. They would keep steady the level of water in the bath. Opinions differ on whether this was possible, or how.  In my opinion, what we saw was more a failure of management than a call for funky new theories.

 

*It is sweet now to remember that there was a phoney war period when this was a potentially contained episode merely about credit, or “the subprime crisis”, and not a devastating collapse of nominal incomes.  That period needs exhaustive study.

 

Advertisements

4 thoughts on “On economics not being weather forecasting

  1. Giles, thanks for your kind words. V small point – when you cut and pasted the von Neumann quote from the PDF, it turned 10^25 into 1025. It is therefore misleading as it stands. I spotted it because when I grabbed the original quote I encountered the same oddity when I pasted! I agree that Cameron/Balls psychology involves much more than just not understanding the intellectual history of economics and what mathematicians/physicists have said about the appropriation of tools from their fields into economics. Cheers, Dom. Ps. In the Endnote of my essay I discuss some interesting (to me) new developments in ‘econophysics’, partly driven from Santa Fe Institute, which may be of interest.

  2. What do scientists do? They form a theory about the causes of a phenomena which exists or could exist, publicise their thoughts and often they, or other scientists, try find out by analyzing data or performing experiments to see if the theory holds. It seems to me that by this definition, economists are scientists. Sure sometimes the theory may claim that the function is so complicated predictions can only be presented statically or perhaps not at all, like quantum mechanics predictions for individual particles. Sometimes scientists have partial theories, they have some parts of the explanation but not all. But that doesn’t make them not scientists.

    I think this debate is really about something else. What people really want is economic engineering, that is an established set of theories about how to run an economy in the “best way”. But as all good engineers know, the first part of engineering is to agree the specifications. This is the really hard part, should we optimize for equality?, or GDP growth, or middle class wages? or long term innovation, or avoidance of crashes? Some or all of these are conflicts, we can’t do all of them at the same time. Until we all agree on what should be optimized, we can’t really engineer an economy. If we did somehow agree on the optimization, then the next question is whether there is the understanding and tools necessary to manage the economy accordingly. I am maybe panglossian but I think we now have a much better understanding than in the past, through the work of such people like Keynes and Friedman about what might be called unforced errors. So I think we can design a system that at least manages that. Future improvement of the system, just like in engineering, is then always possible, but it doesn’t make the overall point go away. One other point, to continue my engineering analogy, a car designer can design a car to survive a crash, but not when, where and if you will crash. And cannot prevent a car from crashing except with some very severe limitations. And there is no point in calling up the designer during the crash to ask him where the car is going to land. It doesn’t mean though he cannot do his job as a car designer. Same with economic crashes, it maybe that the are totally unpreventable, and that a good economic design just to allow for the possibility. It does not mean however that economics engineering has failed.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s