In today's FT, Paul Klemperer argues that those that are sceptical about climate change models should be even more worried than the orthodox concerning the potential dangers of human-induced climate change.
Klemperer's argument is that if we consider standard climate change models subject to great uncertainty, then that means that things might be much worse than the standard models predict, just as much as that things might be better.
Klemperer is mistaken, because he neglects the argument from induction. In situations in which we don’t have reliable out-of-sample models of the future it does not follow that we have high uncertainty concerning the future. For example, I contend that models of neuroscience and psychology are inadequate to predict my wife’s future behaviour with any great certainty. Does that mean I should worry that she might behave in a way radically different from the past – e.g. by poisoning me or stabbing me in my sleep? No. For where we lack robust out-of-sample models, we are ill-placed to predict radical departures from past experience and the appropriate conclusion is that matters will procede much as they have done before – we are not ill-placed to say anything at all.
If our climate models are poor, all that means is that we cannot rely on predictions from them that the earth’s climate will change markedly. It does not follow that we should therefore think it more likely that the world might end.