Daniel Finkelstein is a great Tory writer so it gives me no pleasure to disagree with a recent post of his on Comment Central. Mr Finkelstein rates very highly a book by Nassim Nicholas Taleb, an economist, called The Black Swan, and its supposed relevance to the credit crunch (to be specific, the post is about how to set the pay of bankers in the light of the credit crunch). I think he's wrong to believe that the book (which seems only tangentially related to the famous problem of inductive logic, by the way, making the title curious, despite the fact that Taleb explicitly links his book to that problem) offers anything novel. What's more, if a political party ever attempted to implement a policy based on 'Black Swan' thinking, the effects could be ghastly. Here's why.
The Black Swan (the book, not the inductive logic problem) is about improbable events with bad outcomes. Think 'Meteor strikes the Earth'. It's not very likely, but it is possible, and if that event happened, the outcome - the cost - would be, ah, unlikely to be beneficial (unlike the observation of a black swan, whose existence could only fairly be said to have a terrible impact in people who believed the hypothesis "All swans are white". And I suspect they'd get over it). Mr Finkelstein argued in his post that the crunch (or some of its consequences) should be seen in that light; an unlikely event with a terrible cost. Standard tools for predicting outcomes, for designing pay and reward systems, therefore failed. This is, of course, exactly the narrative that the Prime Minister wants us to believe.
However. The credit crunch was not an extremely improbable event, unless we're now to believe that ramping up buckets of cheap, bad debt, and pretending that one's capital balance remains unaffected by the existence of those buckets, while simultaneously making a mess of the regulatory infrastructure, was ever likely to lead to a great future (of course, this was the very soul of Brown's chancellorship: and this is the point we should be making). As ConservativeHome noted yesterday, David Cameron said this some time ago. The credit crunch was neither improbable, nor unpredictable, nor unpredicted. To be fair, Dr Taleb makes this point in Mr Finkelstein's article. The failure was political, not statistical.
But suppose - just for the sake of my argument - that the crunch was an improbable event, so improbable that no rational thinker would suppose it likely to happen in any finite instance of time. Mr Finkelstein urges that some sort of decision-making apparatus is called into existence to deal with such events. But of course such an apparatus already exists: neither the characterisation of the probability distributions for rare events, nor the decision-theoretic calculus required to choose an action to optimise the utility of the decision one will make in the face of that uncertainty, are closed books; certainly not to statisticians, and by extension I would imagine neither to economists, which makes me suspect that the author of the Swan book is trying to surf a zeitgeist, rather than contribute anything novel (no sour grapes there then, Graeme). To imagine otherwise, I have to suppose that the book's author had never considered that statistical tools exist to deal with questions like "Will there be another earthquake around here, anytime soon? How much preventative action should we take based on that chance?". Or that he'd never heard of insurance.
Now here's the danger of basing policy on a perceived certainty of extreme events, rather than seeing those events embedded within a proper probability distribution. Decision Theory is a big topic, but it boils down to calculating the expected cost of your decision. If you think that a bad outcome will cost a billion, but only has a 1/100000000 chance of happening; while a good outcome will cost a million, but has a 1/2 chance of happening ... you can see where I'm going. To act as though the worst will happen with certainty is to lie to yourself about the chance of the event (it clearly isn't certain). Statisticians call this incoherence, a rare example of elegant statistical terminology. It's quite easy to show that not to obey the probability calculus when making a decision - to bet incoherently - will lead to financial ruin, with probability one (this is one of the fundamental results in the discipline, known as the Dutch Book paradox, due to Ramsey and de Finetti).
Acting as though the worst is going to happen, regardless of the chance or cost of that worst happening, is also, I suggest, essentially what the crystal-botherers call 'The Precautionary Principle', and if put into practice would have the effect of causing us to do nothing, ever. I'd really rather not live my life like that, and I certainly wouldn't vote for a party were its economic platform based on such scare-mongering. Of course bad things might happen; of course they will happen sometimes, even if they had low probability: but we deal with them properly through rigorous application of the probability calculus, whose axioms will govern the frequency of any outcomes, regardless of whether or not policy-makers choose to take notice of their existence.
