An assessment of NIESR forecast accuracy

Abstract

The Institute periodically reviews the accuracy of its macroeconomic forecasts. Pain et al. (2001) compare the performance of NIESR output forecasts to a naïve forecast that uses a simple rule to predict growth next year. They find that between 1980 and 2000 the National Institute forecast performed better than a naïve, or random walk, forecast in two years out of three. Poulizac et al. (1996) consider a sequence of quarterly economic forecasts published by NIESR between 1982 and 1995 (beginning with that produced in February for the growth of GDP and inflation in the following year and finishing with the forecast produced in November for growth in the current year). They show how the reliability of the Institute's forecast improves as the forecast horizon approaches and conclude that errors to GDP and inflation forecasts are normally distributed. In a similar vein, Mitchell (2005) shows that the Institute's point forecast of inflation is reliable whilst its measure of uncertainty has been exaggerated.