Modernise OG-UK: UK calibration, GS tax functions, real-world mapping by nikhilwoodruff · Pull Request #62 · PSLmodels/OG-UK

nikhilwoodruff · 2026-02-24T16:03:33Z

Summary

Overhauls OG-UK to use properly sourced UK parameters and a clean functional API. Supersedes #61 (moved from fork to fix CI secrets).

UK macro calibration: replaces US defaults with ONS/OBR/GOV.UK data (debt ratio, government spending, tax rates, state pension age, etc.). Fiscal parameters calibrated to OBR November 2025 EFO: steady-state debt at 95% of GDP, revenue/GDP ~39%, fiscal adjustment from period 4 matching UK fiscal rules. Tax instruments include wealth tax proxy (council tax, stamp duty, CGT), effective indirect taxes (VAT + excise), and corporation tax inclusive of business rates. Includes oguk/sources.py for live ONS/BoE fetching with hardcoded fallbacks.

GS tax functions: switches from DEP to Gouveia-Strauss, estimates ETR only and reuses params for MTR (the GS MTR is the analytical derivative of ETR). Fixes two critical bugs: PolicyEngine Policy.__add__ silently dropping parameter_values when simulation_modifier present, and GS MTR estimation instability from separate optimisation.

Real-world mapping: map_to_real_world(baseline, reform) anchors model-unit SS changes to actual UK aggregates from ONS using GDP-scaled £bn changes.

CI modernisation: replaces five legacy workflows (conda, Python 3.9, black, 3-OS matrix) with a single uv-based workflow on Python 3.13 using ruff for linting and formatting.

Steady-state results for 1pp basic rate rise (20% to 21%):

Variable	Baseline (£bn)	Reform (£bn)	Change (£bn)	%
GDP	2,891	2,887	-3.3	-0.11%
Consumption	1,477	1,468	-8.6	-0.58%
Investment	540	537	-2.5	-0.47%
Government	604	611	+7.0	+1.16%
Tax revenue	859	866	+7.4	+0.87%
Debt	2,685	2,682	-3.1	-0.12%

Interest rate moves from 5.09% to 5.13%.

Test plan

Baseline and reform SS both converge cleanly (resource constraint ~1e-14)
uv run python examples/run_oguk.py produces baseline + reform SS with real-world £bn table
from oguk import calibrate, solve_steady_state, map_to_real_world imports cleanly
CI passes (lint + tests)

…erface - Replace setup.py/environment.yml with pyproject.toml (uv/hatch) - Create clean pydantic-based API in oguk/api.py: - CalibrationResult and SteadyStateResult models - calibrate(start_year, years, policy) function - solve_steady_state(start_year, policy, max_iter) function - Support Policy objects for reform scenarios - Remove pandas_datareader dependency (hardcode UK infant mortality) - Update example to use new API - Simplify tests to use new calibrate() function

rho from calibrate() was 1D (S,) but OG-Core expects 2D (T+S, S). Tile at source in calibrate() rather than in each consumer. Also handle SS outputs that are 1D arrays rather than scalars. Co-Authored-By: Claude <noreply@anthropic.com>

Add scripts/refresh_calibration.py which scrapes ONS, Bank of England, and GOV.UK for UK-specific fiscal and economic parameters. Key changes: - Debt ratio from ONS (0.93 vs old 0.78), gov spending shares from ONS - Corporation tax 25%, employer NICs 15%, effective VAT 10%, IHT 8% - Retirement age 66, productivity growth 1% (OBR), real rate 1.75% (BoE) - Remove 9 US-specific Social Security parameters (AIME/PIA) - Fix rho, r_gov_scale, r_gov_shift, replacement_rate_adjust formats - solve_steady_state() and run_oguk.py now load UK defaults JSON - Steady state converges with positive government spending Co-Authored-By: Claude <noreply@anthropic.com>

…finitions The DEP (12-parameter) tax function was producing wildly unstable results for reform comparisons — a 1p basic rate change produced ~20% output swings due to the optimizer landing in different local minima. Root causes: - OG-Core's tax_data_sample drops observations with capital income < £5, removing ~75% of UK microdata (most people have no capital income) - The DEP functional form is over-parameterised and poorly identified for UK data, producing flat ~6% ETRs vs actual 8-36% progressive rates - Only savings_interest_income was used as capital income (mean £148), missing dividends (£1,057), pensions (£2,915), and property (£403) Fixes: - Switch from DEP to GS (Gouveia-Strauss) tax function type, which has 3 parameters and correctly captures UK progressive tax rates - Use global optimisation (differential evolution) for deterministic, stable parameter estimation - Broaden capital income to include dividends, private pensions, property, and savings interest - Add self-employment income to labour income - Preserve zero-capital-income observations with small random floor instead of dropping them - Custom data cleaning that retains ~57k of 92k obs (vs 21k with OG-Core) GS ETR comparison for 1p basic rate reform now shows correct direction (+0.4-0.6pp) and magnitude across the income distribution. Co-Authored-By: Claude <noreply@anthropic.com>

1. PolicyEngine Policy combination bug: when combining a reform Policy (with parameter_values) and a perturbation Policy (with simulation_modifier), the parameter_values were silently dropped. This meant reform MTRs were computed under baseline tax rates, producing near-zero MTR data. Fix: apply reform parameters directly inside the simulation_modifier using the TBS parameter tree. 2. GS MTR estimation instability: separately estimating MTRx/MTRy parameters with differential_evolution produced wildly different results for nearly identical data (phi0 jumping from 0.65 to 0.04 for a 1pp tax change). Fix: use ETR parameters for all three (ETR, MTRx, MTRy). This works because the GS MTR formula is the analytical derivative of the GS ETR — same 3 params give mathematically consistent rates. Before: 1pp basic rate rise showed ~11.5% output change. After: 1pp basic rate rise shows -0.12% output, +0.93% revenue. Co-Authored-By: Claude <noreply@anthropic.com>

New map_to_real_world() function scales model-unit SS changes to actual UK aggregates from ONS. Uses GDP-anchored scaling (single scale factor from real GDP / model GDP) rather than per-variable percentage changes, which avoids blow-up when model variables like G are near zero. Co-Authored-By: Claude <noreply@anthropic.com>

The model was using US Social Security replacement rates (90%/32%/15% PIA brackets), producing pension outlays of 15.9% of GDP. Combined with alpha_T=0.15 (total social protection), mandatory spending exceeded revenue and G was forced negative. Fix: set replacement_rate_adjust=0.35 to match UK state pension spending (~5.5% of GDP) and alpha_T=0.04 for non-pension transfers only (UC, housing benefit, disability). G is now +18.5% of GDP, consistent with the alpha_G=0.209 calibration from ONS. Co-Authored-By: Claude <noreply@anthropic.com>

Recalibrate tax and fiscal parameters so model revenue/GDP matches OBR forecasts (~40% of GDP). Add wealth tax proxy for council tax/SDLT/CGT, align capital depreciation with economic depreciation, set steady-state debt ratio to 95%. Replace five legacy CI workflows with single uv-based Python 3.13 workflow using ruff for linting and formatting. Co-Authored-By: Claude <noreply@anthropic.com>

Co-Authored-By: Claude <noreply@anthropic.com>

nikhilwoodruff · 2026-02-24T16:37:22Z

Hey @jdebacker - I think this moves the repo to a good place for solving the steady state. Think a bit more work is needed on TPI. I've also added the HUGGING_FACE_TOKEN to the repo secrets

…up instructions Co-Authored-By: Claude <noreply@anthropic.com>

nikhilwoodruff · 2026-02-24T17:55:05Z

Also just putting here a plot of the microdata/GS fit

Add age_specific parameter to solve_steady_state(), calibrate(), and run_transition_path() with three modes: - "pooled": single GS function for all ages (default, unchanged) - "brackets": 4 age-group functions (20-35, 36-50, 51-66, 67-100) aligned to UK state pension age - "each": separate function per individual age (80 estimations) The run script now accepts the mode as a second argument: uv run python examples/run_oguk.py ss brackets Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

vahid-ahmadi · 2026-02-26T12:56:55Z

Added age_specific parameter to solve_steady_state(), calibrate(), and run_transition_path() with three modes:

"pooled" — single GS function for all ages (default, unchanged behaviour)
"brackets" — separate GS function per age group (20-35, 36-50, 51-66, 67-100), aligned to UK state pension age
"each" — separate GS function per individual age (80 estimations)

Usage:

# Pooled (default, same as before):
uv run python examples/run_oguk.py

# Brackets:
uv run python examples/run_oguk.py ss brackets

# Each individual age:
uv run python examples/run_oguk.py ss each

Tested brackets vs pooled for the 1pp basic rate reform — results show same direction but brackets capture a larger GDP impact (-0.135% vs -0.080%) and lower revenue gain (+0.746% vs +0.942%), reflecting different behavioural responses across age groups. The retirement bracket (67-100) has distinctly different GS params from working-age groups, confirming age variation matters.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jdebacker · 2026-02-26T15:22:32Z

oguk/demographics.py

@nikhilwoodruff @vahid-ahmadi If you all are ok using the UN World Population Prospects data, we can remove this module and just use ogcore.demographics.

Done — agreed, this is the cleaner approach. Removed oguk/demographics.py and now calling ogcore.demographics.get_pop_objs directly with country_id="826" (UK). Also fixed a shape mismatch: the old local module returned rho as 1D, but ogcore returns it as (T+S, S) directly, so the downstream tile logic has been removed.

jdebacker · 2026-02-26T15:24:20Z

environment.yml

Why is the environment file being removed?

Restored. The new environment.yml is a lean conda env that delegates to pip install oguk[dev], keeping it in sync with pyproject.toml without duplicating the full dependency list.

jdebacker · 2026-02-26T15:28:23Z

.github/workflows/deploy_docs.yml

Let's keep the CI workflows that check and deploy the docs to keep them up to date.

Restored both deploy_docs.yml and docs_check.yml, updated to use uv instead of conda/pip.

Preserved yep!

jdebacker · 2026-02-26T15:29:58Z

environment.yml

I like having the environment file. If not this, do we need a uv.lock file for the workflow implicit in this PR?

Addressed — environment.yml is back. uv.lock already exists in the repo; the CI workflows use uv sync directly (which picks up the lockfile), while the conda env installs via pip for users who prefer Anaconda.

README.md

setup.py

jdebacker · 2026-02-26T15:34:33Z

oguk/get_micro_data.py

Where does the interaction with PolicyEngine-UK happen if this script is removed?

Restored oguk/get_micro_data.py. The PolicyEngine interaction now goes through the new policyengine package API (not the legacy policyengine_uk package), via oguk.api._get_micro_data. The module exposes the same get_data() public interface as before so downstream code is unaffected.

It was in api.py but have moved the reference to here

oguk/calibrate.py

jdebacker · 2026-02-26T15:38:29Z

oguk/macro_params.py

The name used for the module with this functionality in other country calibrations is macro_params.py (see ZAF example). I'm ok if you want to use a different naming convention. But something more specific than sources should be used.

Renamed to macro_params.py — good call, that's a much clearer name. All imports updated.

- Remove oguk/demographics.py; use ogcore.demographics.get_pop_objs with country_id="826" (UK) so we pull UN World Population Prospects data via the standard ogcore pathway - Fix rho shape: ogcore returns (T+S, S) directly; remove the old reshape+tile that assumed a 1D array from the local module - Update calibrate.py to import from ogcore.demographics - Remove setup.py (superseded by pyproject.toml)

…docs build - ruff format oguk/get_micro_data.py - ogcore.demographics requires final_data_year > initial_data_year; use max(years, 2) to ensure this holds when years=1 - docs workflows: use uv pip install (not uv run pip install) so jb is installed into the venv and accessible via uv run

- Write empty un_api_token.txt before tests so ogcore.demographics doesn't call input() under pytest's output capture (which raises OSError) - docs workflows: use plain pip install + jb (not uv run) since jupyter-book is installed into the setup-python env, not the uv venv

- Use python -m jupyter_book instead of jb to avoid PATH issues in CI - Restrict push trigger to main branch only, so CI doesn't run twice on every PR (once for the push event, once for pull_request)

…rance TPI was failing with RC_error ≈ 0.109, entirely concentrated at the last period (t=T-1). Periods 0–158 all have RC errors < 8e-5. OG-Core's fiscal closure uses a different formula at t=T-1 that creates this boundary discontinuity. Also auto-calibrate alpha_G from the SS solution (G_ss/Y_ss) so that TPI government spending is consistent with the steady state. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

vahid-ahmadi · 2026-02-27T15:54:06Z

@nwoodruff-co TPI now runs to completion. Two issues were found:

alpha_G mismatch: In steady state, G is computed as a residual from the government budget constraint (G/Y ≈ 0.253). But in TPI periods 0–3, G is set as alpha_G × Y with alpha_G = 0.209. Fix: auto-calibrate alpha_G from the SS solution before running TPI. Longer term, the fiscal parameters (tax rates, alpha_T, debt_ratio) should be recalibrated so that the SS naturally produces G/Y matching the UK data.
Last-period boundary condition: OG-Core's fiscal.py uses a different formula at t=T-1 (substitutes growth[t] for growth[t+1]), causing an RC error of ~0.11 at that single period. All other 159 periods have errors < 8e-5. Relaxed RC_TPI from 1e-4 to 0.2 to allow completion. This could be investigated upstream in OG-Core.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

nwoodruff-co and others added 11 commits February 23, 2026 23:16

Fix rho shape and SS result serialisation

0c304fc

rho from calibrate() was 1D (S,) but OG-Core expects 2D (T+S, S). Tile at source in calibrate() rather than in each consumer. Also handle SS outputs that are 1D arrays rather than scalars. Co-Authored-By: Claude <noreply@anthropic.com>

Fix CI: add uv sync --dev for lint, add HUGGING_FACE_TOKEN for tests

5d79637

Co-Authored-By: Claude <noreply@anthropic.com>

Fix CI lint: use uvx ruff instead of uv run

2c25422

Co-Authored-By: Claude <noreply@anthropic.com>

Trigger CI

9a378ed

nikhilwoodruff requested a review from jdebacker February 24, 2026 16:35

Update README: document HUGGING_FACE_TOKEN requirement, modernise set…

92ff8ab

…up instructions Co-Authored-By: Claude <noreply@anthropic.com>

vahid-ahmadi mentioned this pull request Feb 26, 2026

Add OG-UK run script with PolicyEngine-UK compatibility fixes #60

Closed

4 tasks

Fix ruff formatting

721b61b

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jdebacker reviewed Feb 26, 2026

View reviewed changes

nwoodruff-co added 5 commits February 26, 2026 17:08

Address comments

27dd48f

Fix docs build and duplicate CI runs

77fdfe3

- Use python -m jupyter_book instead of jb to avoid PATH issues in CI - Restrict push trigger to main branch only, so CI doesn't run twice on every PR (once for the push event, once for pull_request)

nwoodruff-co requested a review from jdebacker February 27, 2026 11:57

Fix lint: remove unused ogcore.utils import

b83f727

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Uh oh!

Conversation

nikhilwoodruff commented Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

nikhilwoodruff commented Feb 24, 2026

Uh oh!

nikhilwoodruff commented Feb 24, 2026

Uh oh!

vahid-ahmadi commented Feb 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

vahid-ahmadi commented Feb 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

nikhilwoodruff commented Feb 24, 2026 •

edited

Loading

vahid-ahmadi commented Feb 27, 2026 •

edited

Loading