“It’s better to roughly right than precisely wrong” - John Maynard Keynes
Medieval mathematicians would not have understood the concept of having a one-half chance of getting tails when tossing a die. It’s tails, they might have said, because an omnipotent being wills it. The creation, or rather let’s say the discovery, of probability changed that view.
Outcomes, from a roulette table to a house catching on fire, started to be considered as one of many random possibilities. This not only led to statistical techniques that could show which medical trails are better off but spun out a new form of thought, an element of control. It wasn’t Nietzsche who killed God, it was P(Tails when tossing a coin) = ½.
The techniques used in economic research are wonderfully beautiful and ridiculously complex. These models generally show us that if we do this, then we get that - something along the lines of “if we reduce the higher rate of tax by 2%, this will stimulate growth by 10.384% and increase GDP by a third of a billion”. Fantastic predictions albeit usually wrong ones with spurious accuracy.
It might sound like undermining my own pile of certificates, but the more complex models are, the more likely it is for their predictions to go haywire, as they depend on more assumptions and input parameters.
Take for example two models with two inputs: an additive and multiplicative one. If the inputs are 2 & 2, the outcome is 4 for any model (2+2 = 2x2). What if the input parameters should have been 1.8 and 1.9? A multiplication gives 3.42 while a sum gives 3.8. The more complex model tends to higher fluctuations. In technical jargon, it is considered too “sensitive”.
There are ways of diminishing sensitivity to these minute changes, and significant time should be spent on testing assumptions (and the data) but the argument still holds. Complex models have also not been able to predict extreme scenarios: ranging from clustering of storms in Europe to ruin of financial contracts.
Better believe it
The problem of predictions is exacerbated by two extremes – the (few) experts in the field, and the numerically illiterate. The former establish assumptions set in stone and use them to forecast something. This process changes economics, which has an element of social science, into a pure scientific subject. The results are reassuring as they are mathematical sound, resulting in a form of circular reasoning.
Yet the numerically illiterate are also to blame for wrong decisions, as they do not form what-if questions. They either believe these results without a doubt or simply do not. In Malta, this may be accentuated with blindly following what the political party belief is.
As experts’ predictions continuously do not materialise, mistrust in them increases. This in turn creates more extreme camps and possibly crowds out independent thinkers, putting democracy at greater risk.
My suggestion is: believe and distrust predictions in equal measure, and always ask whether the results make sense.
Dominic believes himself to be God’s gift to the world. Then again does God exist? Contact Dominic on Twitter @domcortis