Learning from forecast errors: the Bank’s enhanced approach to forecast evaluation

By Raphael Abiry, James Hurley, Paul Labonne, David Latto, Harry Li, Andre Moreira, Joseph Oyegoke and Sumer Singh

The Bernanke Review and forecasting challenges of recent years have highlighted the importance of continuous learning from forecast errors. Following substantial investment by Bank staff, and alongside the publication of a new Forecast Evaluation Report in 2026, this paper provides technical detail on the Bank's enhanced forecast evaluation approach. We describe a wide range of evaluation techniques allowing us to characterise the Bank's forecast performance statistically, as well as to interrogate specific forecast errors in greater detail, including their economic drivers. Worked examples are provided throughout, focusing on a subset of macroeconomic variables relevant to monetary policymakers. The statistical techniques described in this paper are also implemented and published as part of a new Python package, to facilitate ongoing forecast evaluation and continued development of this toolkit.

Learning from forecast errors: the Bank's enhanced approach to forecast evaluation