Imagine a world where the climate models are better than the covid ones

As we all know economists only make predictions to prove they’ve got a sense of humour. There is though one thing that economists do in fact get right about their models - probability.

Fraser Nelson over at the Spectator finds that the covid modellers at Sage haven’t grasped this. Those Sage models and results are not adjusted for the probability of outcome, they are instead simply the results from extreme ends of the spectrum of possible outcomes.

We do see exactly the same thing in much of the hyperventilating about climate change. Results from the RCP 8.5 model (A1 FI in the earlier SRES) are what drive all those tales of imminent droughts, London underwater (at the same time too) and the immediate turning of the Earth into Venus. They’re also wrong, in that we know that RCP 8.5 isn’t going to happen and never really was either.

However, we also see folks getting this right even in climate change. Back some years James Hansen decided to model what the carbon tax should be and came up with $1,000 a tonne CO2-e. Immediately folks pointed out that this was not true. His calculation was that if everything went wrong, economic growth was slow, globalisation retreated, we didn’t frack and so turned back to coal as our prime energy source, there were significant positive feedbacks we don’t know about, then maybe and possibly the correct carbon tax could be as much as $1,000 a tonne. The if and the could being very important qualifiers there.

When we include probability - what is the probability of each of the socioeconomic scenarios which drive emissions estimates - then we come back to our more usual range of what that carbon tax should be. As with all other such investigations, ranging from a negative tax - a subsidy to the glories of fossil fuels which make us richer - through to that extreme at the other end, the $1,000. Multiply each outcome by the probability, average out and we get the range we actually use for decision making, somewhere in the $30 to $100 range. As shown by the economists, Nordhaus and Stern.

The estimate must be the averaging of the outcome times the probability. This economists know. This at least some of the discussion about climate change acknowledges. This Sage and covid are entirely ignoring.

Which is why covid modelling is doing such a sterling job in making even economic and climate change estimates look good.

Something of a pity that this basic truth about modelling and prediction is being ignored in the one of the three that requires immediate and significant decision making, isn‘t it?