The Bank of England published Staff Working Paper No. 1,143 setting out a “Blockwise Boosted Inflation Model” (BBIM) that combines boosted decision trees with an open-economy hybrid Phillips curve structure to decompose UK Consumer Prices Index (CPI) inflation into interpretable demand, supply and trend contributions. Applied to monthly UK CPI inflation, the model indicates the post-pandemic inflation surge was driven primarily by global supply shocks transmitted through supply chains, with tight labour markets amplifying inflationary pressures. The approach uses monotonicity constraints to separate demand- and supply-consistent effects and identifies an L-shaped Phillips curve relationship, where inflation responds more sharply when labour markets are very tight; earlier episodes show non-linearities more closely associated with broader slack during recessions. The paper also finds short-term household inflation expectations have exhibited persistent non-linear effects, temporarily lifting trend inflation and prolonging pressures, while longer-term expectations remain anchored. In forecasting exercises, BBIM performs competitively versus linear benchmarks and unstructured machine learning approaches, including reported reductions in root mean squared errors of 10–25% relative to a simple autoregressive model. As a staff working paper, it is presented as research in progress to elicit comments and debate and does not represent Bank of England policy.
Bank of England 2025-09-26
Bank of England staff paper proposes Blockwise Boosted Inflation Model and links recent UK inflation surge mainly to global supply shocks
The Bank of England's Staff Working Paper No. 1,143 introduces a "Blockwise Boosted Inflation Model" (BBIM) analyzing UK Consumer Prices Index inflation by separating demand, supply, and trend contributions. It attributes the post-pandemic inflation surge mainly to global supply shocks and tight labour markets, with short-term household inflation expectations temporarily elevating trend inflation. BBIM demonstrates competitive forecasting performance, outperforming linear benchmarks and unstructured machine learning models.