Simple outperforms complex

With stock markets around the world close to all-time highs, can we declare the global financial crisis ended? If so, when? What is a financial crisis anyway?

My suggestion is that a ‘crisis’ is when a threshold of individual actors realise that the recently successful ‘simple decision rules’ no longer apply – and panic follows. Three examples of ‘rules of thumb’ that suddenly proved alarmingly false are

  • AAA rated means very low probability of default, and very low loss given default’, or
  • US house prices across the whole country are unlikely to fall’.

or (from the 1970s crisis)

  • Inflation is high in the UK, the pound is being devalued, hence property can only rise in value’.

When untested assumptions such as these suddenly come under scrutiny, herding behaviour, the conventional ‘safety in numbers’ approach, becomes hazardous to survival as everyone rushes for the same exit. That is not to say rules of thumb are irrational: if one of our ancestors in the primeval forest sees a fellow tribesman running in the opposite direction, he doesn’t wait to find out what is chasing them.  The crisis ends when these simple rules are jettisoned, only to be replaced by other, equally fallible, rules: ‘The cost of governments borrowing will remain low indefinitely’?

Indeed, simple decision rules have actually made a comeback into regulators thinking. In the Bank of England’s Financial Stability Paper No. 28 (May 2014), Taking uncertainty seriously: simplicity versus complexity in the financial system, the authors point out where simple rules might outperform complex ones when regulating the financial system.  The paper is a collaboration between the Bank and the Max Planck Institute Berlin, and it argues that simple indicators often outperformed more complex metrics in predicting individual bank failure during the crisis. For instance, that the greater complexity of risk weighting embedded in Basel II may sometimes lead to worse performance in comparison to total assets.

As the report says: “Required capital is ‘high’ when there is recent memory of high default rates. But it falls sharply when tranquillity resumes, memories fade, and high default rates drop out of the look-back period used to estimate probabilities of default.”  Specifically, four UK banks that failed or needed rescuing – Alliance & Leicester, Bradford and Bingley, HBOS and Northern Rock – had above-average BIS ‘Tier 1 ratios’.

My own experience of ‘simple outperforms complex’ comes from June 2008, when, as an equity analyst covering banks, I used a basic model built in a couple of hours one Friday after lunch in the pub. I worked backwards, building an ‘embedded value’ type of model used by insurance companies to show the assumptions necessary for Bradford and Bingley’s equity to be worthless. I then published the research note, with a zero pence price target on Bradford and Bingley.

This raised a few eyebrows, to say the least. On the following Monday morning I received a phone call from the Bank of England who asked me to send through my model.

Emma Murphy (who is also one of the authors of the BoE collaboration with the Max Planck Institute) later told me that it was a very powerful model, because it was so simple it was possible to change assumptions and see clearly how much stress the UK mortgage  bank could withstand.

A diverse set of academics (Spyros Makridakis, Robyn Dawes, Dan Goldstein and Gerd Gigerenzer) have all found that, albeit in different contexts, when the environment is highly uncertain, simple models work better than complex ones.

This of course raises a question: ‘when are more complex models superior?’  The answer, according to Gerd Gigerenzer in Gut Feelings: Short Cuts to Better Decision Making (2008), is that “complex analysis, by contrast, pays when one has to explain the past, when the future is highly predictable.”

As if to support Gigerenzer’s assertion, more than a year after I had published my model, a firm of accountants was paid by the government to value Bradford and Bingley equity. Long after the bank had failed, these accountants published a detailed and complex report saying that they believed the equity was worthless, a conclusion for which, though it was highly predictable, they were paid over a million pounds.