Sunday, 4 November 2012

538, Knowledge and Other Things I Don't Understand.

Apparently there’s an election on in the US. You may be expecting a portentous sub-broadsheet editorial to follow, explaining why in spite of past failings this young, middle class student with a twitter account is reluctantly endorsing Barack Obama. But I’m not going to do that, after all, comment is free and you get what you pay for. Rather than anything so vulgar as paying attention to candidate’s policies or personal qualities, I want to take a look at a far less important battle. This is the current argument over election predictions and the reliability of modelling.

538, run by Nate Silver, burst onto the scene in 2008 by predictably predicting the predictable election with predictable accuracy. This was followed by rather less impressive results in the 2010 midterms and a disastrous attempt at fathoming British politics, but this election will be the real test. His model analyses state by state polling, running endless simulations of the election to give approximate percentage chances of each candidate winning. The results haven’t been encouraging for Mitt Romney, who has never really got above 40%. Even at the height of ‘mittmentum’ following the first debate, the model came out 61/39 in Obama’s favour, and since then, the race has got ever less interesting. Hurricane Sandy has punned Romney away, with his chances edging downwards past the 20% mark.

Fittingly, the reactions of both sides to these forecasts have been predictable. The left are overjoyed at each new day’s news, though whether this is triumphalism or self-reassurance is another matter. Bloggers announce the death of political punditry, opinion and personality replaced by numbers and algorithms. The right, meanwhile, rather than attempt to rebut the methods, have gone straight for Silver himself, with the number of attack pieces increasing by the day. They accuse him of favouring Obama, using biased models to drive the narrative and stifle any momentum Mitt Romney may have been building.

On the face of it, it seems that how much confidence someone has in 538 pretty much comes down to whether they want it to be true. This probably accounts for most posts on the subject, but is it whole story? If the positions were reversed, would the right still dismiss such forecasts as witchcraft? Is there something about the underpinnings of the left’s ideology that disposes them to put faith in such models?

Consider intellectuals on the right (yes, they exist). In his Nobel Prize acceptance speech, rapper and Austrian economist FA Hayek railed against ‘the pretence of knowledge’, the fallacy that we could ever have enough data to fully predict the behaviour of the market.  That is not to say that economics is totally futile, we can model how perfectly rational actors would behave in controlled circumstances, but these actors and circumstances do not exist. Any economic predictions for the future of the real world are never more than guesses. Similarly, Nassim Taleb in ‘The Black Swan’ wrote about the futility of attempting to model the future and the potential for completely unpredictable events to change everything. Helpfully for this post (and unhelpfully for his intellectual credibility) he endorsed Austrian torchbearer Ron Paul earlier this year, placing him firmly on the right.

In light of this scepticism, they believe planning by government is doomed to failure, and that individuals, privy to information about their unique circumstances, are better off making decisions for themselves. It is simple to see how this can be expanded from economics to politics, and a rejection of the whole idea of mathematically predicting an election. More recently, conservatives in the US have turned sharply to the right. To win over the tea party grassroots, candidates are embracing increasingly radical views. Could Austrian scepticism of empirical evidence in economics spill over into a total rejection of science as a whole, tying in with the fanaticism of the religious right?

On the other hand, in attempting to plan an economy, the left require predictions not only to be made, but also to be true. The volume of numbers necessary may be immense, but with enough data it is possible for the state to drive the economy in any way it chooses, be it for the benefit of the many or the few. The 538 question therefore becomes a proxy for a bigger one, which may be considered existential for the left. If we can’t accurately predict the results of a binary election, what hope do we have of predicting the future of the country? How do we know what consequences government policies will have? And if we can’t predict the consequences of government policies, how can it ever hope to serve the people?

Since the demise of Marxism as a credible political ideology, this Keynesianism has been the only game in town for the mainstream left, which has in many cases become increasingly pragmatic and managerialist in its philosophy. This has extended into large pushes aimed at 'evidence based policy', superficially devoid of ideology altogether. In areas such as drugs, public health and even welfare, to make a moral argument is awfully 20th century.

To cast the entirety of modern politics as rationalism versus empiricism in this way is clearly stupid. Obama inherited his Keynesian stimulus policies directly from Bush Jnr, and Romney would do little more than attempt to manage the economy to benefit the rich. Nevertheless, this may shine a light onto the thought processes of their supporters, and it is worth considering arguments like that over 538 in a little more depth.

Of course, even if we take the predictive power of Nate Silver’s model at face value, there is an obvious problem. Readers on both sides see an 80/20 split as ‘Obama will win’. Though this is a perfectly reasonable assumption, if we run the election five times Romney will likely steal once. Sooner or later it’s going to be catastrophically wrong, it’s just a question of how much goodwill he can store up before than happens. And if he somehow gets it right every time? Well why bother with an election at all?