Emulator of the OBR macroeconometric model
OBR emulator
Run it hosted, from the CLI or locally, and pick a lever.
Shock a variable, read the deviation
Start hosted: the MCP server
Connect to the hosted MCP server: no install, no credentials, three tools.
The three tools3 points
obr_shockshocks one variable in model units and returns the per-quarter GDP deviation.list_reform_variableslists the shockable levers with units.score_reformwithmodel="obr"scores a PolicyEngine reform end to end through the static-costing bridge (how a reform reaches it).
Or from the CLI
The pe-macro CLI mirrors the hosted tools one for one.
pe-macro compare runs the same reform through
the static and the macro route side by side.
pe-macro obr-shock --var CGG --shock 1250
pe-macro variables
pe-macro score --country uk \
--reform '{"gov.hmrc.income_tax.rates.uk[0].rate":0.21}' \
--model obrInstall and run locally
The package installs from GitHub. The equation listing, the October 2026 EFO forecast tables and the ONS snapshot ship inside it, so there is no data step.
pip install git+https://github.com/PolicyEngine/obr-macroeconomic-modelfrom obr_macro import run_reform
# £5bn/yr government spending shock (£1.25bn per quarter)
results = run_reform(
name="Spending rise",
var="CGG", # government consumption (exogenous)
shock=1250, # £1.25bn per quarter, in £m
periods=12,
)
print(results[["period", "delta_gdp_bn", "pct_gdp"]])
# ΔGDP ≈ £1.27bn rising to £1.30bn — the shock mostly enters the GDP
# identity, so the multiplier is about 1.01 to 1.04 (OBR: 0.6);
# borrowing rises by the cash cost (nominal_consistent=True, the default)About this stepnotes
A reform is one call: pick an exogenous variable, apply a shock, read the deviation from the anchored October 2026 baseline (why the deviation is clean). It is two solves: about one to five minutes for a raw shock on the hosted server.
python -m obr_macro.reform_analysis runs the five
bundled scenarios, with charts landing in outputs/: a £5bn spending rise, a 5pp corporation-tax cut
and rise, a £10bn spending cut, and £10bn public real investment (~£3bn nominal
CGIPS per quarter, which does not reach GDP under the default closure).
To score a statutory reform end to end from Python, install the
suite instead (pip install "policyengine-macro[models] @ git+https://github.com/PolicyEngine/macro#subdirectory=integration")
and call policyengine_macro.score_reform(country="uk", reform={...}, model="obr").
Pick a lever
Reforms enter through the 253 exogenous inputs; results come out of the 383 endogenous variables in the published listing. About 290 of its 372 equations are live in a solve; a variable without a live equation passes through at the OBR’s value.
About this stepnotes
CGGGovernment consumption — spending shocks, £m per quarterCGIPSCentral government investment — £m per quarter (nominal). No transmission by default; investment_closure=True passes it through as an identity, not a multiplier (limit)VRECVAT receipts — £m per quarter (2,200 is a 1pp rise). Barely reaches GDP by default; price_passthrough=True is a labelled overlay (limit)TCPROCorporation tax rate — rate, e.g. −0.05 for a 5pp cut. Converges slowly: a 12-quarter run captures only ~40% of the full effect, and the result reports its plateau_fractionHHDI_ADDFACTORExternally costed household reform — virtual instrument, quarterly £m; positive revenue lowers HHDI via a held add-factorR / GILT / RXBank Rate · gilt yields · exchange rate — monetary & market assumptionsGDPMGDP at market prices — headline outputCONSPrivate consumptionIBUS / IFBusiness / total investment — IF has no live equation under the default closure and moves exactly zeroLFSURUnemployment rate — rate, pp. Does not respond to any lever tested: no Okun channelAll 636 codes, with descriptions, ONS series and equations, are in the model package's variable and equation references. Before quoting a result, check the limits: only income tax and NICs carry behaviour.