We examine the forecasting record of HM Treasury for GDP and the RPI from 1971 to the present. As well as presenting the usual statistical measures of performance, such as Root Mean Squared Errors, and regression tests of forecast efficiency and bias, we test for any relationship between the errors in GDP and RPI forecasts. Confidence intervals are constructed using a classical statistical approach based on past forecast errors, which is similar to that employed in this Review to describe forecast uncertainty.