Dogs and Complexity
One of the more thoughtful regulators around is Andrew Haldane of the Bank of England whose speech "The Dog and the Frisbee" from 2012 remains the touchstone for anyone wanting to appreciate the reasons that modern economics has made a mess out of understanding the real world. To boil the whole thing down to a single statement: you can't control a complex system with complex rules, complex systems require simple rules.
Applied to regulation this is revolutionary, applied to investing it adds to reams of evidence suggesting that most investors need to avoid overcomplicating their analyses, that simple rules of thumb outperform detailed analysis and that mostly we'd be better off going off to the beach with a Frisbee rather than poring over the latest numbers. It's all guesswork anyway, best take the easy option.
Haldane's point is that dogs don't catch Frisbees by solving complex equations and adjusting for gravity, but they succeed in doing so anyway despite the fact that it's a fiendishly difficult mathematical problem. In fact they use a simple heuristic:
"Run at a speed so that the angle of gaze to the Frisbee remains roughly constant. Humans follow an identical rule of thumb."
If, instead, they - or we - were to try and solve the physics behind the problem we'd never catch the damn thing. But the point is, of course, that a dog can catch the Frisbee even though in mathematical modelling terms it appears next to impossible and no one believes that the hound is calculating anything very much. Everyone, that is, apart from economists and regulators who have been beating a determined path away from the idea that simple rules can produce good results where complex analysis all-too-often fails.
Although, to be honest, I doubt many economists believe that this is true any more. The trouble is that the discourse of economics, the idea of the rational, self-serving individual leading to the best allocation of capital is now endemic. It's spread beyond the confines of economics into every nook and cranny of the world, infiltrating thought patterns and infecting institutions: so any attempt to change this from within is more than welcome.
Haldane argues that most modern regulation has been about managing risk, following the two key economic models of decision making under risk - the Arrow-Debreu framework, is based on the cardinal assumption of equilibrium (discussed in Exit the Walras, Followed By Equilibrium and sundry other articles), and the portfolio allocation framework developed first by Harry Markowitz and elaborated by Robert Merton - which uses past results to estimate future ones (see, for example, Markowitz's Portfolio Theory and the Efficient Frontier). Both models ignore uncertainty, and as such both models fail because they assume that people are rational.
And this is why the world of modern finance is so complicated - because it's full of experts who are so busy using these models to try and figure out how to catch Frisbees they continually run over the edge of cliffs. And like Wile. E. Coyote it often takes them a while to figure this out.
Out of this discussion Haldane deduces five commandments for decision making under uncertainty which we can apply directly to investors.
- Firstly, thou shalt use simple decision rules in a complex environment.
- Secondly, thou shalt recognize that ignorance is bliss.
- Thirdly, thou shalt understand that past results are a fragile guide to the future.
- Fourthly, thou shalt appreciate that complex models need huge data sets.
- Fifthly, thou shalt avoid complex rules for they shall cause you to act like a mindless box-ticker.
Admittedly that's not quite what Haldane says, but I think it captures the main points. To take them one at a time.
Thou shalt use simple decision rules in a complex environment.
You won't improve your returns by doing more and more detailed analysis of numbers: we'll run out of computing power before we catch the Frisbee. Stick to simple asset allocation rules, embrace diversification and filter on straightforward numbers that anyone can understand.
Thou shalt recognise that ignorance is bliss.
It's counterintuitive but disregarding information can produce better results. Gerd Gigerenzer and Stephanie Kurzenhäuser in Fast and Frugal Heuristics in Medical Decision Making showed that a simple decision tree outperformed both physicians' intuition and the existing complex decision support tool (see: It's How Big, Not How Often, That Counts). Or: should an investor rely on their gut or a clever statistical model? (Answer: neither).
Thirdly, thou shalt understand that past results are a fragile guide to the future.
Despite copious amounts of data, analysts and economists remain completely unable to predict market turning points. As Haldane points out simple tallying strategies are far more predictive of avalanches than more complex detection methods. We saw something similar in The Turkey Illusion: An Audience With Gerd Gigerenzer where a simple 1/N asset allocation strategy (where N is the number of available asset classes) outperformed more complex models.
Fourthly, thou shalt appreciate that complex models need huge data sets.
This is a point we've seen before. In theory, with enough data and enough computing power a complex model can outperform simple heuristics - short-range weather forecasting for example (although I often have doubts about that). But for stockmarket prediction that requires far more data than we'll ever have.
Fifthly, thou shalt avoid complex rules for they shall cause you to act like a mindless box-ticker.
Finally, complex rule sets often lead to a mindless application of the rules or defensive decision making. People follow the rules prescriptively, either because it's easier than thinking for themselves or because they fear the consequences of not doing so. Gigerenzer and Kurzenhäuser describe this effect on physicians: if you don't follow the rules and someone dies you get sued, even if the person would have died anyway.
The tendency of economists and analysts to overcomplicate things is down to a combination of things. Partly it's egotistical - if you've spent years learning how to model complex systems you're hardly going to accept that you can be outperformed by a Border collie. Partly it's man-with-a hammer syndrome - if the only tool you have is complex modelling then that's what you'll damn well use.
Similar issues can occur with investors, who are all too inclined to go to one extreme or the other. At the far end of the spectrum there are lots of incredibly complicated models that pertain to predict the future trajectory of stocks, while at the other there's a descent into following the nearest intelligent sounding person or the use of mystical incantations like charts, gurus and tipsheets.
These latter approaches are misapplications of heuristics, of course: herding is a sensible social rule in an unfamiliar environment and mysticism is a reaction to lack of control in uncertain environments. Neither is appropriate to investing: it's important to pick the right heuristic for the right model, otherwise it's just man-with-a-hammer syndrome by another route.