Bank of England structural VAR

BoE SVAR

How the replication and the forecasts score, and where it departs from the Bank’s model.

GDP variance share consistent with the paper; no forecast skill

Two kinds of check: replication against the published paper, and forecasts scored against outturns.

What was checked, against what, and the result
CheckAgainstResultVerdict
Global shocks’ share of UK GDP varianceBrignone and Piffer (2025): ~40%, 4 quarters38.7% (68% band 24.0–53.1)pass
Global shocks’ share of UK CPI variancePaper: ~50%, 4 quarters41.5% (25.6–56.3), 8.5pp shortpartial
Domestic shares and their rankingPaper: ~40% for both; monetary policy the largest38.6% / 36.8%; monetary policy 8.9% of CPI variance, third of threepartial
Exact identitiesZero and sign restrictions, revision adding-up; gated in CI; software, not economicsmax error 3.64e-12by construction
One frozen forecastONS outturns 2024Q3–2026Q1, from the 2024Q2 edge; one originRMSE 0.32pp GDP and CPI; 14 of 14 in the 68% band (about 9.5 expected)partial
Point-forecast accuracyRandom walk with drift; 50 origins, 8 variables × 8 horizons0 of 64 survive adjustment (smallest q 0.39)fail
Band coverageNominal 68% and 90%; 50 origins; mean of 8 variables, CPI on the published year-on-year basis68% band 70% → 57%; UK CPI 90% band 78% → 54%weak
Monetary policy shock sizeBank of England (1999): 0.2–0.35% of GDP for a 1pp riseabout 1.6% of GDP per 100bp at h = 8 (pre-refresh run, not recounted)fail

Evidence by dimension

Evidence by dimension: the registry’s categorical judgement, and what would change it
DimensionLevelWhat would change it
Implementation fidelitystrongAll exact invariants hard-gated for every spec
Predictive validationweakDensity scores and rolling coverage, post-Covid re-run
IdentificationmoderateResults shown across lag, prior and proxy grids
Policy counterfactualsnot applicableKeeps refusing reform scoring
UncertaintymoderateRolling coverage and proper predictive scores
Vintage reproducibilitymoderateFrozen manifests; real-time and revised vintages
About this replicationIndependent; not a Bank of England product

An independent replication built from Bank of England Macro Technical Paper No. 3 and public data. It is not produced, maintained or endorsed by the Bank of England; do not present its results as Bank of England estimates.

The four tests

Data and sampleData through 2026Q2, sample from 1992Q1

Every figure here uses data through 2026Q2, except test 2 (frozen origin, scored through 2026Q1), and the estimation sample 1992Q1–2025Q1 (boe-var-model #24); earlier figures are at commit 845e2a4.

What CI gates, and what is refreshed by hand
  • Gated in CI: the replication invariants and the forecast archive's append-only rule.
  • Computed once and refreshed by hand: the 50-origin rolling evaluation and the coverage study.
  • The sign and zero restrictions are imposed by the identification scheme, so that gate protects the software and cannot be evidence about the economy.
1

Against the Bank of England paper

GDP share 38.7% vs ~40%; figures 2–6 match qualitatively

Global shocks’ variance shares, against Brignone & Piffer (2025).

Check by check against Brignone & Piffer (2025)
CheckThis replicationThe paper
Global shocks' share of UK GDP variance (4 quarters) 38.7% [24.0, 53.1] 68% ~40%
Global shocks' share of UK CPI variance (4 quarters) 41.5% [25.6, 56.3] 68% — about 8.5pp short ~50%
Domestic shocks' share, UK GDP / UK CPI (4 quarters) 38.6% / 36.8% “around 40%” for both
Identified shocks' total share, UK GDP / UK CPI 77.3% / 78.3% “around 80%”
IRFs, FEVDs, shocks, historical decompositions Signs and shapes match Figures 2–6, on an earlier run; magnitudes not yet checked (#11) qualitative match
Forecast-revision adding-up identity exact per draw (max abs error 3.64e-12) holds by construction — a regression check, not an agreement

Share of forecast-error variance from global shocks, %

BoE SVAR: global-shock FEVD shares, ours vs Brignone & Piffer (2025) Grouped bar chart. Share of UK forecast-error variance attributed to identified global shocks (world demand, energy and supply) four quarters ahead, as a share of total variance — the statistic the paper's Figure 4 plots. For GDP, our production artifact gives 38.7% against the paper's 40.0%, 1.3 points short; for CPI, 41.5% against 50.0%, 8.5 points short. Neither gap is resolvable: on the 600-draw check configuration the 68 per cent posterior band on the same share runs 24 to 57 per cent for GDP and 24 to 54 per cent for CPI, both far wider than the shortfall, and the paper's values are read off its figure and approximate. This comparison and the fan chart elsewhere on this page are different runs: the fan chart is the frozen 2024Q2-edge run of 3,000 draws at seed 20250717.
  • ours
  • paper
Global-shock (world demand + energy + supply) FEVD shares four quarters ahead: the production run (10,000 posterior draws, 765 accepted, ESS 368.0; results/summary.md at boe-var-model c827ab8) against the paper's approximate published values.

Source: BoE SVAR replication

Source details

Source: papers/boe-svar/figures/comparison_numbers.json.

How noisy that comparison is

The 68% posterior band is about ±15pp wide, so the 8.5pp CPI shortfall sits inside a single interval: a real gap on the paper's definition, reported as one, but not resolvable at this sample size.

The CI gate uses a cheap ~50-draw configuration whose global GDP share moves about ±6.5pp across sampling seeds, checked against a wide [30, 60]% band. It catches gross regressions; it is not a second opinion on the production run.

About this stepthe definition of the share

The comparison uses the paper's own definition: the posterior mean of the per-draw group share of total four-quarter-ahead variance, which leaves about a fifth unexplained.

2

One frozen origin, scored against outturns

RMSE 0.32pp; 14 of 14 in band is over-dispersion

Seven quarters from the 2024Q2 edge, against the ONS outturns since.

% change on a year earlier

BoE SVAR: out-of-sample forecast fan from the frozen 2024Q2 edge vs ONS outturns Two-panel fan chart. Left panel: year-on-year UK GDP growth; right panel: year-on-year UK CPI inflation. Each shows the posterior median forecast from the frozen 2024Q2 data edge as a line, the 68 per cent credible band as a shaded region over thirteen quarters 2024Q3 to 2027Q3, and ONS outturns for the seven evaluated quarters 2024Q3 to 2026Q1 as dots. All fourteen outturn dots fall inside the 68 per cent band; RMSE 0.32 percentage points on GDP and 0.32 on CPI. GDP medians run 1.3, 1.8, 1.3, 0.9, 0.8, 0.9, 0.9 per cent over the evaluated quarters against outturns of 1.0, 1.5, 1.3, 1.4, 1.3, 1.0, 0.9; CPI medians 2.3, 2.7, 2.8, 3.2, 3.2, 3.2, 3.2 against outturns of 2.0, 2.5, 2.8, 3.5, 3.8, 3.4, 3.1. A dashed rule in each panel marks 2026Q1, the last quarter scored (the test was not re-scored on the 2026Q2 refresh): the six quarters to its right are forecast with nothing yet to check them against, so the fit shown covers seven of the thirteen quarters plotted. This run is 3,000 posterior draws at seed 20250717, 217 of which were accepted by the sign restrictions; it is a different run from the FEVD comparison elsewhere on this page, which comes from the production artifact. Medians and 68 per cent bands from papers/boe-svar/figures/figure_numbers.json (forecast_table, entries [median, lo68, hi68]); ONS outturns from papers/boe-svar/figures/make_figures.py. Coordinates: GDP panel maps value v to y = 292 - (v + 1) * 59 for the -1 to 3 per cent axis; CPI panel y = 292 - (v - 1) * 59 for the 1 to 5 per cent axis; quarter i of 13 maps to x = 58 + i * 25.667 (GDP) or 416 + i * 25.667 (CPI).
  • median forecast
  • 68% band
  • ONS outturn
  • last quarter scored (2026Q1)

Each panel starts at the frozen 2024Q2 edge; scored through 2026Q1 and not re-scored on the 2026Q2 refresh; the 6 quarters right of the dashed rule are not evaluated.

Every outturn sits inside the 68% band, which is over-dispersion, not a pass. Frozen 2024Q2 data edge; source: papers/boe-svar/figures/figure_numbers.json.

The model put inflation at about 3.2% from mid-2025 and held it there. Outturns peaked up to 0.7pp above that (3.5 and 3.8% in 2025), inside the 68% band but on its upper half, and had fallen to 3.1% by 2026Q1. It never saw the Ofgem cap increases or the April 2025 administered price rises.

Seven scored quarters

Window
2024Q3–2026Q1
A correct 68% band
holds about 9.5 of 14
2025 CPI outturns
3.5% and 3.8%
Medians then
3.2%
Peak miss
up to 0.7pp
About this stepout-of-sample · why 14 of 14 is not a pass

Out-of-sample. Seven quarters ahead from the frozen 2024Q2 edge, the RMSE in the scorecard above holds for both GDP growth and CPI inflation.

Why 14 of 14 is not a pass. A correct 68% interval should contain about 9.5 of 14, and one frozen origin yields only two or three effectively independent observations. Weight the 50-origin coverage instead, which finds the opposite problem.

Full quarterly scorecard: working paper.

3

Across 50 origins

68% band covers 70% at h=1, 57% at h=8

The rolling evaluation: the reading to weight over the single origin.

Share of outturns inside the band, %

BoE SVAR: empirical band coverage across 50 origins Line chart of empirical interval coverage by forecast horizon, averaged across the eight model variables, against the nominal 68 and 90 percent levels; a vertical bar at each horizon spans the best and worst variable. 68% band, mean coverage at horizons one to eight: 70%, 70%, 64%, 61%, 60%, 61%, 60%, 57%; 90% band, mean coverage at horizons one to eight: 86%, 83%, 81%, 78%, 77%, 74%, 74%, 72%. The 68 per cent band covers 70% at one quarter, just above nominal, and falls below nominal from horizon 3 onward to 57% at eight; the 90 per cent band under-covers at every horizon, from 86% down to 72%. The mean hides how bad the worst variable is: the lowest cell in the grid is CPI energy at the 68 per cent band, horizon 4, covering 30%. The window includes the Covid quarters and the evaluation model carries no Covid dummies, which depresses coverage.
  • 68% band, mean coverage across variables
  • 90% band
  • nominal coverage

Vertical bars span the best and worst of the eight variables; lowest cell 30% (CPI energy, 68% band, h4).

Coverage across 50 expanding-window origins, mean of the eight variables, by horizon. The 68% band over-covers at one quarter (70%) and falls below nominal from three onward; the 90% band under-covers at every horizon.

Source: BoE SVAR replication

Source details

Source: papers/boe-svar/figures/coverage_evaluation.json.

50 expanding-window origins

Run
evaluation model without Covid dummies
68% band, h=1
70%, just above nominal
68% band, h=8
57%
90% band
under-covers at every horizon
Unemployment satellite
ahead of no-change at h=1–4 only excluding Covid quarters, not significantly; bands cover less than nominal
About this stepthe runs behind the chart · the unemployment satellite

The runs behind the chart

  • Coverage worsens with horizon: the opposite of the frozen-origin reading, and the one to weight.
  • The chart is the evaluation run, without Covid dummies.
  • With the six Covid dummies (2020Q1–2021Q2) that the published fans carry, the 68% band covers about the same: 70% at one quarter and 56% at eight, or 59% at eight scored on year-on-year growth, the basis of the published fans.
  • UK CPI alone, 90% band, same specification and basis: 78% at one quarter, 54% at eight (boe-var-model production_spec).
  • Each origin uses 1,000 draws and the Cholesky shock factor, so coverage is the same on every machine (boe-var-model #26, #30).
  • No variable beats a drifting random walk at any horizon once the 64 tests are adjusted together (test 4); three attempts to recalibrate the bands after the fact are in the limits.

The svar-unemployment satellite, over 73 origins at h=1–4 (67 targets a horizon once the Covid quarters are excluded; specification on the methodology tab; unemployment_satellite_validation.json):

  • Point forecast: RMSE relative to no-change 0.82–0.99 excluding the Covid quarters, not significant; on all 73 targets 0.99 at h=1, then 1.12–1.35. The published path is capped at four quarters.
  • Bands in the archived 28 Jul 2026 round (GDP uncertainty only): 22–48% of outturns inside the 68% band, 30–61% inside the 90% band.
  • Bands the model now produces, adding the satellite’s residual and estimation uncertainty (boe-var-model #32): 60–66% at 68%, below nominal at every horizon; 87%, 78%, 75% and 70% at 90%, close to nominal only one quarter ahead. On all 73 targets: 58–63%, and 81%, 75%, 74% and 68%.
  • The backtest draws its GDP paths from a refit without Covid dummies, not the published specification; no round with the wider bands is archived yet.
4

Against a drifting random walk

No win survives multiple testing; Bank Rate closest

Error ratios for eight variables at horizons 1–8, across the same 50 origins.

Benchmark diagnostics, and what official forecasters report

BoE SVAR against a random walk, computed from our own runs

Expanding-window design: 50 quarterly origins (2012Q1–2024Q2), estimation from 1992Q1, data through 2026Q2, horizons 1–8. A ratio below 1.0 beats the benchmark.

The benchmark decides the answer.

  • A no-change walk on a trending log level forfeits the whole trend as error.
  • So beating it on a price index is close to uninformative.
  • Against it, 3 of 64 tests survive adjustment at 5% in the published specification, all one quarter ahead: world GDP (0.96), world CPI (0.74) and UK CPI (0.62), each at q = 0.048. None does in the evaluation run, and none against drift in either.
  • UK CPI's 0.63 at one quarter becomes 0.83 against a walk with drift.
  • At eight quarters, 0.67 becomes 1.03.
  • Bank Rate is the one series that does not trend.
  • It is also the one that improves under the harder benchmark.

Covid dummies matter for GDP, not for inflation.

  • The rolling evaluation omits the six Covid dummies (2020Q1–2021Q2).
  • Every published forecast here carries those dummies.
  • Published specification: UK GDP against drift goes 1.06 → 0.99 at one quarter.
  • At eight quarters it goes 1.12 → 0.95.
  • CPI and Bank Rate move by less than 0.02.
  • Level with a naive benchmark is still not beating it.
  • UK GDP is statistically inseparable from drift at every horizon (p = 0.33–0.43).
  • That result holds in both specifications.

The AR(1) comparison in the source file is not usable at long horizons: 94.6% of its eight-step squared error for UK GDP comes from one origin, because an AR(1) extrapolates the Covid collapse. Every benchmark carries a worst_origin_mse_share diagnostic for that reason.

What the OBR and external forecasters report about themselves

Cited, not recomputed: the OBR's Forecast Evaluation Report (July 2025) reports its own real-time errors on annual growth rates since 2010, beside the HM Treasury median of external forecasters.

Median absolute forecast errors reported in the OBR Forecast Evaluation Report, July 2025 (annual rates, percentage points, forecasts since 2010)
Variable · horizonOBRExternal median
Real GDP growth · 1 year ahead0.6pp0.6pp
Real GDP growth · 2 years ahead0.4pp0.4pp
CPI inflation · 1 year ahead0.3pp0.3pp
CPI inflation · 2 years ahead0.9pp0.9pp

The nearest in-house analogue is the frozen-edge RMSE above.

  • It is a different statistic, frequency and origin count.
  • So it sits beside these numbers, not against them.
  • For the Bank of England the only sourceable comparison is a matched-vintage episode.
  • The BoE MPR (Aug 2024) put late-2025 CPI at ~2.4% modal.
  • The outturn was 3.8%, a ~1.4pp miss.
  • The MPR publishes no RMSE-by-horizon table.
  • OBR FER (July 2025): one-year GDP growth forecasts under by 0.4pp on average.
Omitted for lack of a defensible source: NIESR or consensus RMSE tables by horizon, a Bank of England RMSE-by-horizon table, and an AR(1) benchmark for this exercise. The authors' companion paper (Staff Working Paper No. 1,165) is a methods paper on a different SVAR and publishes no errors, so the frozen-edge RMSEs stand without an official yardstick.

RMSE ÷ drifting-random-walk RMSE, horizons 1–8 quarters

BoE SVAR forecast skill against a random walk with drift, all eight variables Dot matrix of forecast error relative to a random walk with drift, eight variables by eight quarterly horizons, from 50 expanding-window origins. Position alone carries better or worse: a dot left of the 1.0 rule means the model beats the benchmark, right of it means the benchmark wins. Filled dots mark differences significant at 5 per cent by a Diebold-Mariano test, hollow dots differences that are not statistically distinguishable; those p-values are pairwise and the grid runs 64 tests, none of which clears a 10 per cent false-discovery rate once adjusted together — the smallest adjusted q-value is 0.39. Bank Rate runs 0.79 at one quarter to 0.85 at 8; UK CPI runs 0.83 at one quarter to 1.03 at 8; World CPI runs 0.94 at one quarter to 1.07 at 8; Oil price runs 1.01 at one quarter to 0.98 at 8; CPI energy runs 0.97 at one quarter to 1.11 at 8; UK real GDP runs 1.06 at one quarter to 1.12 at 8; World GDP runs 1.07 at one quarter to 1.11 at 8; Exchange rate runs 1.03 at one quarter to 1.33 at 8. Bank Rate is the best at the longest horizon and Exchange rate the worst. Against this harder benchmark only four of eight variables beat naive at one quarter and two at 8. The number at the end of each row is that row's ratio at horizon 8.
  • significant at 5%, pairwise
  • not distinguishable

Left of the 1.0 rule the model is better; right of it the benchmark is.
The large dot, and the number after each name, is h=8. Smallest adjusted q over the 64 cells: 0.39.

Error ratios against a random walk with drift, on the level of each series — the fair benchmark for a trending series. Estimated without the Covid dummies the published specification carries; no-change ratios are in the table below. Estimation uses final revised data, so this is pseudo- rather than real-time out-of-sample. The ratios rest on fewer effective observations than 50 origins suggests: in 16 of the 64 cells one origin contributes over 35% of the squared error, peaking at 64% for UK GDP at h=1.

Source: BoE SVAR replication

Source details

Source: papers/boe-svar/figures/rolling_evaluation.json (worst_origin_mse_share.drift).

Variables (of 8) with lower RMSE than a random walk with drift

How many of eight variables beat a drifting random walk, by horizon Column chart, one column per forecast horizon from one to eight quarters. Column height is the number of the model's eight forecast variables whose root mean squared error is below that of a random walk with drift; a dashed reference line across the top marks all eight. The counts are 4, 4, 3, 2, 3, 3, 2, 2, so between 2 and 4 of the 8 beat the benchmark at any horizon and the remaining 4, 4, 5, 6, 5, 5, 6, 6 do not. The shaded lower part of each column is the subset whose difference is significant at 5 per cent on an unadjusted pairwise Diebold-Mariano test: 2, 1, 1, 1, 1, 0, 0, 0. Those p-values are pairwise and the grid runs 64 tests; under a Benjamini-Hochberg adjustment none of them reaches a 10 per cent false-discovery rate and the smallest adjusted q-value is 0.39, so on the adjusted reading the significant-win count is zero at every horizon. At horizon 8 one variable is significantly worse than the benchmark.
  • beats it, p<0.05 unadjusted (count under each column)
  • beats it, not distinguishable

Column height = variables with lower RMSE than a random walk with drift, out of 8; the rest do not beat it.
Adjusted together, none of the 64 tests clears a 10% false-discovery rate (smallest q = 0.39).

A win is a ratio strictly below 1.0 — a hard cut that counts 0.99 and rejects 1.01 though neither is distinguishable from the benchmark, so the significance split matters more than the count. That significance is pairwise: under a Benjamini–Hochberg adjustment the significant-win count is zero at every horizon. Same origins and benchmark as above.

Source: BoE SVAR replication

Source details

Source: papers/boe-svar/figures/rolling_evaluation.json.

Forecast error relative to two naive benchmarks (below 1.0 beats it). Bold marks a significant win, * any significant difference — unadjusted, so read a star as "worth a second look". Ratios omit the Covid dummies the published specification carries (see the diagnostics above). The ex-Covid column is against no-change — UK GDP's 0.74 there is 1.20 against drift — and neither it nor the full-sample figure is preferred.
variableh=1h=2h=4h=8h=1 ex-Covid
(vs no-change)
Vs driftVs no-changeVs driftVs no-changeVs driftVs no-changeVs driftVs no-change
Bank Rate0.79*0.880.86*0.960.89*1.020.851.030.86
UK CPI (level)0.83*0.630.850.610.940.661.030.670.62
World CPI (level)0.940.730.950.681.010.691.070.670.73
Oil price1.011.031.011.041.001.050.981.081.05
CPI energy0.970.980.970.981.031.041.11*1.130.97
UK real GDP (level)1.061.061.091.081.101.081.121.050.74
World GDP (level)1.071.041.101.021.110.921.110.700.51
Exchange rate1.031.041.091.111.191.221.331.411.05

These are quarterly, level-basis ratios computed with hindsight-final data. They are not comparable with the real-time annual-growth errors the OBR publishes, which is why those sit in a separate table (in the diagnostics above).

Ratio to drift, h=1

Bank Rate
0.79, the closest to a win
UK CPI
0.83 (0.63 vs no-change)
UK real GDP
1.06
Smallest p
0.019, unadjusted
Adjusted wins
zero at every horizon
About this stepthree statistics, not ranked · the closest to a win

Three statistics answer three different questions and are not averaged or ranked: this SVAR's pseudo-out-of-sample accuracy against naive benchmarks (computed here, quarterly), the OBR's and external forecasters' published real-time errors (cited, annual), and the OBR emulator's tracking of the official forecast (replication, not forecasting).

The closest to a win. Bank Rate at one quarter, 0.79 against a drifting random walk, is the smallest of 64 variable×horizon tests at p = 0.019. Adjusted together (Benjamini–Hochberg), nothing clears a 10% false-discovery rate against the drifting walk on any variable at any horizon.

Where this departs from the Bank’s model

Known limits of the replication
LimitDetail
Frozen data edge Estimation to 2025Q1, conditioning to 2026Q2 — the extended sample, not the paper's 1992Q1–2023Q2 window. Results shift with revisions.
Proxied world aggregates The Bank's UK-trade-weighted world GDP and CPI are unpublished, so they are rebuilt as chain-weighted US + euro-area + Japan + China with time-varying UK trade weights.
One ranking does not replicate
  • The paper’s prose ranks UK monetary policy as the largest domestic contributor to CPI variance.
  • Here UK supply (14.6%) and UK demand (10.5%) both exceed it (8.9%).
  • The paper publishes no FEVD table: unresolved, not a known defect; the proxy aggregates are a candidate cause.
Assumed lag length; loose pandemic prior p = 4 is assumed, not selected; the pandemic is handled by six exogenous Covid dummies with a loose prior, which is the Cascaldi-Garcia (2022) pandemic prior at its uninformative default.
The posterior is mildly explosive
  • In 99.3% of the 10,000 posterior draws the largest root of the VAR is above 1; the median is about 1.004.
  • Only 4 of 744 accepted draws are stable, counted on the pre-refresh run (10,000 draws, 744 accepted, before the 2026Q2 data); not yet recounted on the production run (765 accepted).
  • The cause is the specification (variables in levels, a prior centred on a random walk), not a code error.
  • The paper’s Fig. 3C has the same shape: Bank Rate fades; GDP and CPI levels stay shifted to h20.
  • Beyond h20 the CPI level keeps drifting, an extrapolation the paper never shows; long-horizon responses here carry this (boe-var-model #11).
No inflation anchor
  • With a unit root and no 2% target in the model, the CPI forecast settles near 3.1% (recent inflation), not the target.
  • Re-estimated on 1992Q1–2019Q4 it settles at 2.0%.
  • Read a medium-term CPI gap to the OBR or the Bank as this artefact, not a forecast of overshoot (forecast_vs_boe_obr.md).
No after-the-fact fix brings the 90% band to 90%
  • Scored held-out, the raw 90% band covers about 77% of outturns.
  • Per-horizon rescaling factors: 82.7% (82.9% on year-on-year growth), with a 95% interval of about 68% to 95%, so the gap left to 90% is within sampling error (#19).
  • Conformal factors: 83.2% on the same specification (#20).
  • Lenza–Primiceri Covid volatility scaling: 81.2%, or 86.9% combined with the rescaling (#21).
  • 88% of the misses fall in three episodes, each mostly on one side of the band: 2014Q3–2016Q4, 2020Q1–2021Q2 and 2021Q3–2023Q4. Outside them the raw 90% band covers 94%.
  • So the shortfall is mostly the central forecast missing for several quarters, not tail shape; stochastic-volatility variants, tried and not adopted, left the below-band misses in place (band decision note, #12).
  • Read the 90% fan as under-covering, not as a 90% probability.
Rate-shock response above the Bank of England’s estimate
  • Per 25bp monetary policy shock, GDP is about 0.40% lower at h8 (posterior median, pre-refresh run of 744 accepted draws; not yet recounted).
  • Scaled linearly to 100bp that is about 1.6%, above the Bank of England’s published range for a 1pp rise (the benchmark).
  • The experiments differ: here the rate path is persistent (a unit root), not held for one year, and the GDP level effect does not fade (boe-var-model #11).
Sign-restriction critiques apply Pointwise medians mix structural models (Fry–Pagan); the Haar prior over rotations is informative about impulse responses (Baumeister–Hamilton). These hit the Bank's own outputs identically.

Every deviation is listed in the repository's docs/methodology.md.