Book Review: Nate Silver- The Signal and the Noise

‘Our bias’ argues Nate Silver author of the New York Times blog FiveThirtyEight and now minor American celebrity-nerd ‘is to think that we are better at prediction than we really are’. Thus across a range of fields humans tend to confuse noise for signal; we misinterpret data and produce faulty, if plausible-sounding, forecasts. 
 

The consequences can be grim. An unrealistic model, such as that which grossly underestimated the risk of default on AAA rated securities, can permeate far beyond its intended boundaries, impacting upon society as a whole. There is, therefore, a need to favour accuracy over complexity. Predictions must be put forward with honesty; in other words, they should be presented as estimates ripe for revision, not certainties destined to pass. 

In the first part of the book Silver attempts to illustrate the problem by offering the layman (myself very much included) several handy case-studies. He waltzes from poker to baseball to weather forecasting, explaining how prediction operates in each and how it can both help and hinder. 

These initial chapters also mix in a heavy dollop of autobiography. Silver, before he shot to political polling fame, worked for KPMG and then invented a baseball modelling programme (think Moneyball). These diversions are pleasant enough, and it should be noted that Silver comes across as a likeable, modest character. (What geek doesn’t?) 

The second half is mainly devoted to advocating the merits of the Bayesian model of prediction. I know what you’re thinking, finally! But it is a testament to Silver’s crisp, engaging style that what could be a rather dull topic keeps the readers’ interest throughout. Basically, and I will be very basic, the Bayesian approach factors in new information without sacrificing past actions/trends in order to create a new, updated estimate.

Thankfully, Slver’s explanation is far more elegant and is illustrated via several accessible examples. I won’t dwell on them here though (that’s the job of the book), rather I’ll look at the wider point: that in order to be more accurate (or, depending on your preference, less wrong) predictions need to be flexible; they must adapt to, and incorporate, new information. 

In essence then, Silver is making the case for humility. He is championing not just pragmatism but a mind-set that recognises human frailties and prices them into the predictions which we make. 

By extension he takes issue with those prone to bold statements and those who interpret events as confirmation of their own world view. Such certainty, in Silver’s mind, is evidence of intellectual weakness. Thus, rather belligerent political pundits and market-traders of a certain ilk are both criticised for lacking flexibility.

It seems no coincidence that both strident, ideological politics and brash, swaggering economics figured prominently during the early years of the new millennium. Pushing further (perhaps too far), it is possible to link both these traits to the administration and over-riding philosophy of George W. Bush. 

From this perspective then, The Signal and The Noise champions not just a branch of statistics but also a way of thinking. It celebrates pragmatism, flexibility and a realistic appreciation of what is likely to occur. And given this, I guess it is no wonder that Mr Silver was so sure of President Obama’s re-election.