PolicyEngine Macro

Open models · easy access

Open economic models
for policy analysis.

Five open UK and US models for reforms, forecasts, and economic shocks, with clear assumptions and validation.

Choose a model Get started
boe-svar: UK real GDP growth forecast, median with 68% and 90% credible bands Fan chart. Posterior forecast of UK year-on-year real GDP growth from the boe-svar model, 2026Q2 to 2029Q2, conditioned through 2026Q1. The median path runs from +1.10% to +1.42%. The 90% credible band is widest at 2028Q3, spanning -0.97% to +3.85%, and the lower edge stays below zero throughout the forecast: the model does not rule out a contraction. This figure is the site's hero motif and uses the site's current-forecast artifact rather than the frozen validation run. -1% +0% +1% +2% +3% 2026Q2 2029Q2
Current UK GDP growth forecast from the 2026Q1 data edge, with 68% and 90% uncertainty ranges. Explore the model.
01 — choose by question

Different questions need different models.

Policy decisions affect households and the wider economy over different time horizons. Choose the open-source model that matches your question, then review its evidence and limitations.

Taxes and benefits for a household or populationUK + US · PolicyEngine A monetary or fiscal shock in the US over coming quartersUS · frb-us A selected UK fiscal or macro scenario over 3–5 yearsUK · obr-macro Work, saving, and capital effects over decadesUK · psl-og What drove UK GDP and inflation at the latest model-data vintageUK · boe-svar

Need assumptions and tradeoffs side by side? Compare all five models.

02 — model catalog

Inspect the model before you use it.

Each reference page explains inputs, outputs, methodology, validation, limitations, and how to run the model.

obr-macroUKhosted

OBR macroeconometric model

Run selected UK fiscal and macro scenarios over 3–5 years. Anchored to the March 2026 EFO; borrowing is not yet returned.

read the model →
boe-svarUKhosted

Bank of England structural VAR model

Diagnose the latest model-data vintage and forecast GDP and inflation with uncertainty ranges. Coefficients are estimated through 2025Q1; the model does not score reforms.

read the model →
frb-usUShosted

Federal Reserve macroeconomic model

Simulate selected US shocks under VAR expectations. Hosted raw shocks only; no reform scoring or model-consistent expectations.

read the model →
psl-ogUKlocalexperimental

PSL overlapping generations model

Estimate long-run work, saving, and capital mechanisms. Calibrated research prototype; local solves take at least 17 minutes.

read the model →
pe-microsimUKUShosted

PolicyEngine tax-benefit microsimulation

Calculate UK and US household taxes, benefits, and net income. Deterministic rule calculation; population estimates depend on survey data and do not include macro feedback.

read the model →
03 — validation

Validation is a gradient, not a badge.

We report how closely each model matches an independent benchmark—and say when none exists. The five models span three verification classes; an anchored match is not the same as validation. For example, obr-macro misses by 0.15% when anchored and 5.75% free-running. Compare official benchmarks →

obr-macroPublished anchor

Tracks the March 2026 EFO to 0.15% GDP MAPE.

boe-svarPublished anchor

0.32pp RMSE; 14 of 14 outturns inside the 68% ranges.

frb-usPublished anchor

Matches current pyfrbus 1.1.1 to 1.4×10⁻⁸ on shock responses.

pe-microsimStatute

Deterministic household calculations, checked rule by rule; population estimates add survey uncertainty.

psl-ogCalibrated

Targets met by construction; no replication ground truth.

04 — start using it

Add the models to Claude or ChatGPT—or use them in your own code.

Connect through MCP, then ask Claude or ChatGPT to run a model. Hosted models need no installation or API key. Prefer code? Use the CLI or Python. Connect and try a model →

Claude ChatGPT