Macro Technical Paper No. 2
By Andre Moreira
This paper introduces a Staggered-Combination MIDAS (SC-MIDAS) approach, used by Bank of England staff to nowcast UK GDP and other variables. SC-MIDAS uses a mix of restricted and unrestricted MIDAS regressions and two forecast combination steps to exploit 'hard' and 'soft' data optimally through the release cycle – specifically when the lower-frequency target (quarterly GDP) is also sampled at a higher frequency (monthly GDP). This structure enables it to capture key features of the data, including the mechanical relationship between monthly and quarterly GDP, and to dynamically reweight hard versus soft signals in a way that improves performance compared to standard pooled MIDAS approaches. In practice, SC-MIDAS combines accuracy with interpretability, outperforming several benchmarks out-of-sample and producing a range of outputs that policymakers find useful. Our positive experience with other UK variables shows SC-MIDAS is applicable more widely.
Nowcasting GDP at the Bank of England: a Staggered-Combination MIDAS approach
