The National Institute Density Forecasts of Inflation
Summarised from the National Institute Economic Review, number 193, July 2005. To order a the full version of this article or a subscription, please contact Sage Publications by telephone: +44 (0) 20 7324 8701, email: <a href="mailto:firstname.lastname@example.org">mailto:email@example.com</a> or online at <a href="http://ner.sagepub.com">http://ner.sagepub.com</a>.
The National Institute was one of the first forecasters to publish not just a point forecast of inflation (a forecast of the "central tendency") but also indicate the degree of uncertainty associated with this forecast. This was achieved by publishing uncertainty bands or more recently density forecasts.
This paper examines these uncertainty forecasts. It finds that the National Institute's approach of setting the variance of its forecast equal to the variance of the historical forecast errors needs to be implemented carefully to produce satisfactory uncertainty estimates. Looking too far back into the past, perhaps into a different policy "regime", can generate misleading uncertainty estimates. However, regular monitoring of these historical forecast errors, using recently developed statistical tests for structural breaks at an unknown point, can be used to help select a period of history which is informative about the future.
This paper finds that when forecasting inflation one should not look back beyond 1993; post-1993 inflation has entered a period of "low" volatility. Ahead of this advice since 2002 the National Institute has produced uncertainty forecasts based on examination of forecast errors back to 1993 only. This paper confirms that this is a sensible approach.