Bank of England structural VAR

BoE SVAR

Decomposes UK GDP and inflation into six named shocks, and forecasts both with bands.

GDP variance share consistent with the paper; no forecast skill. Evidence A variance-share comparison with the Bank of England’s paper: global shocks’ share of GDP variance consistent with it, of CPI variance 8.5pp short; no forecast skill against a drifting random walk once the tests are adjusted together.

Good for

  • What is hitting the economy now, as six named shocks
  • GDP and CPI forecast paths with raw, uncalibrated 68% and 90% bands
  • Why the forecast changed since last quarter

Not for

  • Scoring a statutory reform: score_reform does not accept it (limits)
  • Policy counterfactuals: OBR emulator, OG-UK
  • Beating a simple benchmark: no forecast skill (the test)
  • A return-to-target inflation path: no inflation anchor, so medium-term CPI stays near 3.1%, the recent trend (limits)

What it returns

The shocks behind 2026Q2, and the forecast archived on 12 Nov 2026 from the same data.

What hit the UK economy in 2026Q2

World demand: positive95%posterior probability
World energy: negative≥95%posterior probability
World supply: negative70%posterior probability
UK demand: negative95%posterior probability
UK supply: positive75%posterior probability
Monetary policy: contractionary80%posterior probability; the shock-size check fails (why)

Sign: positive = expansionary (raises GDP); energy negative = oil prices up.

From latest_shocks.json: a forecast run of 16 Nov 2026 on data through 2026Q2 (455 of 6,000 draws accepted, ESS 181), not the production estimation run. Probabilities are rounded to the nearest 5%, the precision that sample supports. How the shocks are named: methodology.

GDP growth forecast with bands

UK real GDP, % change on a year earlier

Current BoE SVAR UK real GDP growth forecast Fan chart from the latest complete quarterly data edge, 2026Q2, forecasting 2026Q3 to 2029Q3. Median with 68 and 90 percent predictive bands.
  • median
  • 68% band
  • 90% band

Forecast made 2026-11-12 from data through 2026Q2.

Bands are raw posterior quantiles, not rescaled to measured coverage; the calibrated version, with the CPI fan, is on the forecast record.
Run details

Live-facing forecast, distinct from the frozen 2024Q2 validation experiment. Data through 2026Q2; forecast begins 2026Q3. Estimated on 1992Q1–2025Q1; a separate forecast run, 455 of 6,000 posterior draws accepted, not the production estimation run behind the methodology and validation figures. A separate run from the shock probabilities with the same draw settings, so both show 6,000 draws and 455 accepted. Source: the archived round forecasts/rounds/2026-11-12/boe-svar.json, shown through papers/boe-svar/figures/current_forecast.json.

Every archived round, scored against outturns: the forecast record.

How far to trust it

Three checks, weakest included; each opens its evidence.

Full scorecard and limits

Run it

The shock reading above, three ways, at the same 6,000 draws.

Using the BoE SVAR, which structural shocks hit the UK economy in the latest quarter, and how likely is each sign?The hosted server calls latest_shocks; no install. Connect

pe-macro shocks --draws 6000
from policyengine_macro import svar_latest_shocks
shocks = svar_latest_shocks(draws=6000)
print(shocks["shocks"])

Setup, every lever and output

With the microsimulation

Its CPI path prices the uprating of CPI-linked benefits.

BoE SVARCPI path PolicyEngine microsimulationbenefit uprating Who gains and loses

svar_inflation_incidence prices the gap between the median CPI path and the October 2026 EFO through next April’s uprating of Universal Credit and Child Benefit, by decile. Every join in the suite.

An experimental illustration, not a scored reform. It uses the forecast median and a curated slice of CPI-linked benefits; the state pension and frozen thresholds are excluded.

Model cardinputs, outputs, horizon, data, access
Answers
What is hitting the UK economy now, why the forecast changed, and how uncertain the baseline is
Inputs
Eight packaged quarterly series, all public (ONS, Bank of England, FRED)
Outputs
GDP and CPI forecasts with bands, identified shocks, decompositions, forecast revisions
Horizon
Quarterly short-run forecast
Data
Sample 1992Q1–2025Q1 · conditioning data through 2026Q2
Uncertainty
Posterior 68% and 90% intervals, raw rather than calibrated to measured coverage
Access
Hosted forecast and latest shocks, CLI, Python
Runtime
Minutes per estimation run
Joins the microsimulation
The median CPI path prices benefit uprating (the join above, experimental)
Status
Partial replication of Brignone and Piffer (2025): GDP decomposition consistent with the paper; CPI global-shock share about 8.5pp short; no forecast skill against a drifting random walk
Cannot answer
The effect of a statutory policy reform: score_reform does not accept it (limits)