Forecasts are almost always wrong, we should take them with a pinch of salt,” says Iain Duncan Smith (BBC interview, 12 May). Is he right? Yes.
There are practical and theoretical reasons why economic models are poor predictors. Here are two examples of their failure: the IMF said the UK Coalition government’s austerity policies would cause recession, instead the economy grew better than most; the UK Treasury forecast 2016’s first quarter growth in November 2015 and had to revise its view a few months later. This is normal.
You might think that some forecasters have more authority than others, but power and esteem in the hierarchy matter little. Or you might say that more forecasters favour outcomes on your side of a debate than on the other, but majority opinion is irrelevant.
This doesn’t mean that all models are totally useless. The best they can do is to flag up possible consequences that might happen and would otherwise have been overlooked. Forewarned is forearmed.
If Boris Johnson (or Nigel Farage!) were Chancellor instead of George Osborne, would the Treasury model have given the same predictions about Brexit? Inputs to the model are chosen carefully, therefore we should carefully check who is sponsoring the forecast.
The following links give interesting explanations for the practical shortcomings of economic models (or any models of similarly complex environments):
Here are some of the questions raised in these articles:
How fully and accurately do the equations represent the system?
How accurately can the input values be measured?
- What sorts of events, predictable and potential, could occur that are outside the closed system for which the model has been constructed but could have effects within it?
That last question is the hardest, though the others are hard too. Predictable events should be included in the model, with probability estimates on their likelihood of occurring and on their possible effects on any forecast. That’s difficult enough, which is why the models economists use are invariably closed – to exclude such events. But it’s the unpredictable events that make nonsense of even the best models; where ‘best’ means providing answers to the first three questions.
Gloomy forecasts of the ill effects of Brexit invariably assume – without stating the assumptions – (a) that what will be lost can be credited to the EU, and (b) that it could not be maintained or restored by an independent Britain. In the case of economic forecasts aimed to support the Remain case, the doom forecasters fail to ask whether it would be in the interest of both parties, the EU and Britain, to retain or restore the ‘loss’. It will usually be in both parties’ interest to avoid losing what has been gained.