Introductory article to NIESR's Economic Review, February 2017
There has been an intense debate about the rationale behind economic prediction or forecasting, triggered by a sequence of forecast errors before and after the financial crisis and more recently by a ‘surprisingly’ buoyant economy after last year’s referendum on
the UK’s membership of the European Union. Some economists argue that the value of a forecast is strictly related to its forecast accuracy. Others argue that what matters is less the forecast errors but the stories that are revealed by such errors. The former might be thought
to relate the value of economic forecasting solely in terms of a statistical criterion and the latter to the need to concentrate on structural relationships between economic variables that will be subject to errors (or shocks) but which can be treated as stable. I argue that the forecast process is inherently subject to large errors, and so is a hazardous exercise, but that does not by itself invalidate the exercise because both the producers and consumers of forecasts understand that errors will occur. And this knowledge throws up a clear obligation for producers to explain errors before the fact by use of
uncertainty or scenario plots and for consumers to treat the forecasts with caution.