Merge pull request #163 from PolicyEngine/nikhilwoodruff/issue162

nikhilwoodruff · web-flow · commit 33f379b66e28 · 2025-07-14T16:21:11.000+01:00
Make weights consistent and add 2023 FRS
diff --git a/.github/workflows/pull_request.yaml b/.github/workflows/pull_request.yaml
@@ -60,15 +60,3 @@ jobs:
           path: calibration_log.csv
       - name: Run tests
         run: pytest
-      - name: Test documentation builds
-        run: make documentation
-
-      - name: Check documentation build
-        run: |
-          for notebook in $(find docs/_build/jupyter_execute -name "*.ipynb"); do
-            if grep -q '"output_type": "error"' "$notebook"; then
-              echo "Error found in $notebook"
-              cat "$notebook"
-              exit 1
-            fi
-          done
diff --git a/.github/workflows/push.yaml b/.github/workflows/push.yaml
@@ -68,16 +68,6 @@ jobs:
         run: make upload
         env:
           HUGGING_FACE_TOKEN: ${{ secrets.HUGGING_FACE_TOKEN }}
-      - name: Test documentation builds
-        run: make documentation
-      - name: Build Jupyter Book
-        run: make documentation
-      - name: Deploy documentation
-        uses: JamesIves/github-pages-deploy-action@releases/v3
-        with:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          BRANCH: gh-pages  # The branch the action should deploy to.
-          FOLDER: docs/_build/html
       - name: Publish a git tag
         run: ".github/publish-git-tag.sh || true"
       - name: Remove .whl files
diff --git a/changelog_entry.yaml b/changelog_entry.yaml
@@ -0,0 +1,4 @@
+- bump: patch
+  changes:
+    fixed:
+    - Added calibrated weights from 2022.
diff --git a/policyengine_uk_data/datasets/frs/dwp_frs.py b/policyengine_uk_data/datasets/frs/dwp_frs.py
@@ -108,6 +108,15 @@ class DWP_FRS_2022_23(DWP_FRS):
     time_period = 2022
 
 
+class DWP_FRS_2023_24(DWP_FRS):
+    folder = STORAGE_FOLDER / "frs_2023_24"
+    name = "dwp_frs_2023_24"
+    label = "DWP FRS (2023-24)"
+    file_path = STORAGE_FOLDER / "dwp_frs_2023_24.h5"
+    time_period = 2023
+
+
 if __name__ == "__main__":
     DWP_FRS_2020_21().generate()
     DWP_FRS_2022_23().generate()
+    DWP_FRS_2023_24().generate()
diff --git a/policyengine_uk_data/datasets/frs/enhanced_frs.py b/policyengine_uk_data/datasets/frs/enhanced_frs.py
@@ -1,7 +1,10 @@
 from policyengine_core.data import Dataset
 from policyengine_uk_data.utils.imputations import *
 from policyengine_uk_data.storage import STORAGE_FOLDER
-from policyengine_uk_data.datasets.frs.extended_frs import ExtendedFRS_2022_23
+from policyengine_uk_data.datasets.frs.extended_frs import (
+    ExtendedFRS_2022_23,
+    ExtendedFRS_2023_24,
+)
 from policyengine_uk_data.datasets.frs.frs import FRS_2022_23
 from policyengine_uk_data.utils.loss import create_target_matrix
 
@@ -156,6 +159,14 @@ class EnhancedFRS_2022_23(EnhancedFRS):
     end_year = 2028
 
 
+class EnhancedFRS_2023_24(EnhancedFRS):
+    name = "enhanced_frs_2023_24"
+    label = "Enhanced FRS (2023-24)"
+    file_path = STORAGE_FOLDER / "enhanced_frs_2023_24.h5"
+    data_format = Dataset.TIME_PERIOD_ARRAYS
+    input_frs = ExtendedFRS_2023_24
+    time_period = 2023
+
+
 if __name__ == "__main__":
-    ReweightedFRS_2022_23().generate()
-    EnhancedFRS_2022_23().generate()
+    EnhancedFRS_2023_24().generate()
diff --git a/policyengine_uk_data/datasets/frs/extended_frs.py b/policyengine_uk_data/datasets/frs/extended_frs.py
@@ -2,7 +2,7 @@
 from policyengine_uk_data.utils.imputations import *
 from policyengine_uk_data.storage import STORAGE_FOLDER
 from typing import Type
-from policyengine_uk_data.datasets.frs.frs import FRS_2022_23
+from policyengine_uk_data.datasets.frs.frs import FRS_2022_23, FRS_2023_24
 from tqdm import tqdm
 
 
@@ -103,6 +103,15 @@ class ExtendedFRS_2022_23(ExtendedFRS):
     time_period = 2022
 
 
+class ExtendedFRS_2023_24(ExtendedFRS):
+    name = "extended_frs_2023_24"
+    label = "Extended FRS (2023-24)"
+    file_path = STORAGE_FOLDER / "extended_frs_2023_24.h5"
+    data_format = Dataset.TIME_PERIOD_ARRAYS
+    input_frs = FRS_2023_24
+    time_period = 2023
+
+
 def create_public_services_inputs(sim) -> pd.DataFrame:
     variables = [
         "age",
@@ -184,4 +193,4 @@ def add_public_services(data: dict, simulation, time_period: int):
 
 
 if __name__ == "__main__":
-    ExtendedFRS_2022_23().generate()
+    ExtendedFRS_2023_24().generate()
diff --git a/policyengine_uk_data/datasets/frs/frs.py b/policyengine_uk_data/datasets/frs/frs.py
@@ -190,6 +190,14 @@ class FRS_2022_23(FRS):
     time_period = 2022
 
 
+class FRS_2023_24(FRS):
+    dwp_frs = DWP_FRS_2023_24
+    name = "frs_2023_24"
+    label = "FRS (2023-24)"
+    file_path = STORAGE_FOLDER / "frs_2023_24.h5"
+    time_period = 2023
+
+
 def add_id_variables(frs: h5py.File, person: DataFrame, household: DataFrame):
     """Adds ID variables and weights.
 
@@ -874,7 +882,7 @@ def add_expenses(
 
 
 def add_benunit_variables(frs: h5py.File, benunit: DataFrame):
-    frs["benunit_rent"] = np.maximum(benunit.BURENT.fillna(0) * 52, 0)
+    pass
 
 
 def impute_brmas(dataset, frs):
@@ -931,3 +939,4 @@ def impute_brmas(dataset, frs):
 if __name__ == "__main__":
     FRS_2020_21().generate()
     FRS_2022_23().generate()
+    FRS_2023_24().generate()
diff --git a/policyengine_uk_data/datasets/frs/local_areas/constituencies/calibrate.py b/policyengine_uk_data/datasets/frs/local_areas/constituencies/calibrate.py
@@ -20,7 +20,7 @@
 )
 from pathlib import Path
 from policyengine_uk_data.storage import STORAGE_FOLDER
-from policyengine_uk_data.datasets import EnhancedFRS_2022_23
+from policyengine_uk_data.datasets import EnhancedFRS_2023_24
 
 FOLDER = Path(__file__).parent
 
@@ -32,14 +32,14 @@ def calibrate(
     overwrite_efrs=True,
 ):
     matrix_, y_, country_mask = create_constituency_target_matrix(
-        EnhancedFRS_2022_23, 2025
+        EnhancedFRS_2023_24, 2025
     )
 
     m_national_, y_national_ = create_national_target_matrix(
-        EnhancedFRS_2022_23, 2025
+        EnhancedFRS_2023_24, 2025
     )
 
-    sim = Microsimulation(dataset=EnhancedFRS_2022_23)
+    sim = Microsimulation(dataset=EnhancedFRS_2023_24)
 
     COUNT_CONSTITUENCIES = 650
 
@@ -181,12 +181,13 @@ def dropout_weights(weights, p):
 
             if overwrite_efrs:
                 with h5py.File(
-                    STORAGE_FOLDER / "enhanced_frs_2022_23.h5", "r+"
+                    STORAGE_FOLDER / "enhanced_frs_2023_24.h5", "r+"
                 ) as f:
-                    if "household_weight/2025" in f:
-                        del f["household_weight/2025"]
+                    if "household_weight/2023" in f:
+                        del f["household_weight/2023"]
                     f.create_dataset(
-                        "household_weight/2025", data=final_weights.sum(axis=0)
+                        "household_weight/2023",
+                        data=final_weights.sum(axis=0) / 1.021,
                     )
         l.backward()
         optimizer.step()
diff --git a/policyengine_uk_data/datasets/frs/local_areas/local_authorities/calibrate.py b/policyengine_uk_data/datasets/frs/local_areas/local_authorities/calibrate.py
@@ -12,21 +12,21 @@
     create_local_authority_target_matrix,
     create_national_target_matrix,
 )
-from policyengine_uk_data.datasets import EnhancedFRS_2022_23
+from policyengine_uk_data.datasets import EnhancedFRS_2023_24
 
 DEVICE = "cpu"
 
 
 def calibrate():
     matrix, y, r = create_local_authority_target_matrix(
-        EnhancedFRS_2022_23, 2025
+        EnhancedFRS_2023_24, 2025
     )
 
     m_national, y_national = create_national_target_matrix(
-        EnhancedFRS_2022_23, 2025
+        EnhancedFRS_2023_24, 2025
     )
 
-    sim = Microsimulation(dataset=EnhancedFRS_2022_23)
+    sim = Microsimulation(dataset=EnhancedFRS_2023_24)
 
     count_local_authority = 360
 
diff --git a/policyengine_uk_data/storage/download_private_prerequisites.py b/policyengine_uk_data/storage/download_private_prerequisites.py
@@ -14,6 +14,7 @@ def extract_zipped_folder(folder):
 FILES = [
     "frs_2020_21.zip",
     "frs_2022_23.zip",
+    "frs_2023_24.zip",
     "lcfs_2021_22.zip",
     "was_2006_20.zip",
     "etb_1977_21.zip",
diff --git a/policyengine_uk_data/tests/microsimulation/README.md b/policyengine_uk_data/tests/microsimulation/README.md
@@ -0,0 +1,126 @@
+# PolicyEngine UK reform impact tests
+
+This directory contains automated tests to monitor when model changes affect the fiscal impacts of policy reforms. It helps ensure that updates to the PolicyEngine UK model don't unexpectedly change reform outcomes.
+
+## Overview
+
+The testing framework consists of three main components:
+
+1. **reforms_config.yaml** - Configuration file containing reform definitions and their expected impacts
+2. **test_reform_impacts.py** - Pytest test suite that verifies reforms produce expected impacts
+3. **update_reform_impacts.py** - Script to update expected impacts when model changes are intentional
+
+## Quick start
+
+### Running tests
+
+To check if current model produces expected reform impacts:
+
+```bash
+pytest test_reform_impacts.py
+```
+
+To run with more detail:
+
+```bash
+pytest test_reform_impacts.py -v
+```
+
+### Updating expected impacts
+
+When model changes are intentional and you need to update the baseline expectations:
+
+```bash
+# Preview changes without updating
+python update_reform_impacts.py --dry-run
+
+# Update the configuration file
+python update_reform_impacts.py
+
+# Update with detailed output
+python update_reform_impacts.py --verbose
+```
+
+## Configuration file structure
+
+The `reforms_config.yaml` file defines each reform with:
+
+```yaml
+reforms:
+  - name: "Description of the reform"
+    expected_impact: 10.5  # Fiscal impact in billions
+    parameters:
+      "gov.some.parameter": 0.25
+      "gov.another.parameter": 1000
+```
+
+## Workflow
+
+### Regular testing
+
+1. Run `pytest policyengine_uk/tests/microsimulation/test_reform_impacts.py` as part of your CI/CD pipeline
+2. Tests will fail if any reform impact differs by more than £0.1 billion from expected
+3. This catches unintended consequences of model changes
+
+### After intentional model changes
+
+1. Run `python policyengine_uk/tests/microsimulation/update_reform_impacts.py --dry-run` to preview impact changes
+2. Review the changes to ensure they're expected
+3. Run `python policyengine_uk/tests/microsimulation/update_reform_impacts.py` to update the configuration
+4. Commit both the model changes and updated `reforms_config.yaml`
+
+### Adding new reforms
+
+1. Add a new entry to `reforms_config.yaml`:
+   ```yaml
+   - name: "Your new reform description"
+     expected_impact: 0.0  # Initial placeholder
+     parameters:
+       "gov.parameter.to.change": new_value
+   ```
+
+2. Run `python policyengine_uk/tests/microsimulation/update_reform_impacts.py` to calculate the actual impact
+3. Run `pytest policyengine_uk/tests/microsimulation/test_reform_impacts.py` to verify the test passes
+
+## Test tolerance
+
+Tests allow for a tolerance of £0.1 billion in fiscal impacts. This accounts for:
+- Minor numerical differences in calculations
+- Small variations from microsimulation sampling
+
+If you need to adjust this tolerance, modify the assertion in `test_reform_impacts.py`.
+
+## Troubleshooting
+
+### Tests failing after model update
+
+1. Check if the changes are expected by running the update script with `--dry-run`
+2. If changes are expected, update the configuration file
+3. If changes are unexpected, investigate the model modifications
+
+### Import errors
+
+Ensure you have the required dependencies:
+```bash
+pip install policyengine-uk pytest pyyaml
+```
+
+### Configuration file not found
+
+The tests expect `reforms_config.yaml` to be in the same directory as the test file. You can specify a different path:
+
+```bash
+python update_reform_impacts.py --config /path/to/reforms_config.yaml
+```
+
+## Example reforms
+
+The configuration includes various UK tax and benefit reforms:
+- Income tax rate changes (basic, higher, additional rates)
+- Personal allowance adjustments
+- National Insurance contribution rates
+- VAT rate changes
+- Universal Credit taper rate modifications
+- Child benefit amount changes
+
+Each reform's fiscal impact is measured as the change in government balance (in billions) for the year 2029.
diff --git a/policyengine_uk_data/tests/microsimulation/reforms_config.yaml b/policyengine_uk_data/tests/microsimulation/reforms_config.yaml
@@ -0,0 +1,33 @@
+reforms:
+- name: Raise basic rate by 1pp
+  expected_impact: 7.7
+  parameters:
+    gov.hmrc.income_tax.rates.uk[0].rate: 0.21
+- name: Raise higher rate by 1pp
+  expected_impact: 4.9
+  parameters:
+    gov.hmrc.income_tax.rates.uk[1].rate: 0.42
+- name: Raise personal allowance by ~800GBP/year
+  expected_impact: 0.8
+  parameters:
+    gov.hmrc.income_tax.allowances.personal_allowance.amount: 13000
+- name: Raise child benefit by 25GBP/week per additional child
+  expected_impact: -1.3
+  parameters:
+    gov.hmrc.child_benefit.amount.additional: 25
+- name: Reduce Universal Credit taper rate to 20%
+  expected_impact: -38.3
+  parameters:
+    gov.dwp.universal_credit.means_test.reduction_rate: 0.2
+- name: Raise Class 1 main employee NICs rate to 10%
+  expected_impact: 12.4
+  parameters:
+    gov.hmrc.national_insurance.class_1.rates.employee.main: 0.1
+- name: Raise VAT standard rate by 2pp
+  expected_impact: 19.5
+  parameters:
+    gov.hmrc.vat.standard_rate: 0.22
+- name: Raise additional rate by 3pp
+  expected_impact: 4.4
+  parameters:
+    gov.hmrc.income_tax.rates.uk[2].rate: 0.48
diff --git a/policyengine_uk_data/tests/microsimulation/test_reform_impacts.py b/policyengine_uk_data/tests/microsimulation/test_reform_impacts.py
diff --git a/policyengine_uk_data/tests/microsimulation/update_reform_impacts.py b/policyengine_uk_data/tests/microsimulation/update_reform_impacts.py
diff --git a/policyengine_uk_data/tests/test_aggregates.py b/policyengine_uk_data/tests/test_aggregates.py
diff --git a/policyengine_uk_data/tests/test_child_limit.py b/policyengine_uk_data/tests/test_child_limit.py
diff --git a/policyengine_uk_data/tests/test_childcare.py b/policyengine_uk_data/tests/test_childcare.py
diff --git a/policyengine_uk_data/utils/imputations/consumption.py b/policyengine_uk_data/utils/imputations/consumption.py

Original file line number	Diff line number	Diff line change
`@@ -12,21 +12,21 @@`
`12`	`12`	`create_local_authority_target_matrix,`
`13`	`13`	`create_national_target_matrix,`
`14`	`14`	`)`
`15`		`-from policyengine_uk_data.datasets import EnhancedFRS_2022_23`
	`15`	`+from policyengine_uk_data.datasets import EnhancedFRS_2023_24`
`16`	`16`
`17`	`17`	`DEVICE = "cpu"`
`18`	`18`
`19`	`19`
`20`	`20`	`def calibrate():`
`21`	`21`	`matrix, y, r = create_local_authority_target_matrix(`
`22`		`- EnhancedFRS_2022_23, 2025`
	`22`	`+ EnhancedFRS_2023_24, 2025`
`23`	`23`	`)`
`24`	`24`
`25`	`25`	`m_national, y_national = create_national_target_matrix(`
`26`		`- EnhancedFRS_2022_23, 2025`
	`26`	`+ EnhancedFRS_2023_24, 2025`
`27`	`27`	`)`
`28`	`28`
`29`		`- sim = Microsimulation(dataset=EnhancedFRS_2022_23)`
	`29`	`+ sim = Microsimulation(dataset=EnhancedFRS_2023_24)`
`30`	`30`
`31`	`31`	`count_local_authority = 360`
`32`	`32`