Re-source the rail ridership index to ORR and project rail usage

changelog.d/1862-rail-ridership-orr.fixed.md

@@ -0,0 +1 @@
+Correct `gov.dft.rail.ridership_index` to ORR Table 1220 passenger journeys (2021-22 is 2.55 and 2024-25 4.46 on a 2020-21 base, not 0.40 and 0.95), hold it flat from 2026-27, and project `rail_usage` by its growth per person, which it was not projected by before.

policyengine_uk/data/uprating_indices.yaml

@@ -159,5 +159,9 @@ gov.economic_assumptions.yoy_growth.obr.private_pension_index:
 - private_pension_income
 gov.economic_assumptions.yoy_growth.obr.lagged_average_earnings:
 - housing_service_charges
+# ORR passenger journeys per person, as household weights grow by population
+# (#1862).
+gov.economic_assumptions.yoy_growth.orr.per_capita.rail_passenger_journeys:
+- rail_usage
 gov.economic_assumptions.yoy_growth.ons.population:
 - household_weight

policyengine_uk/parameters/gov/dft/rail/ridership_index.yaml

@@ -1,30 +1,26 @@
-description: Cumulative index of rail ridership (passenger journeys). Base year 2020 = 1.0. Used for uprating the quantity component of rail subsidy spending.
+description: Index of passenger journeys on Great Britain's railways. Financial year 2020-21 = 1.0.
+# ORR Table 1220, passenger journeys in millions (published 18 June 2026),
+# divided by the 2020-21 figure, 387.885m, rounded to 4 decimals. The
+# financial year starting in April of year t is stored at t-01-01. ORR marks
+# every year from 2020-21 provisional.
+#
+# No official body forecasts passenger journeys, so from 2026-27 the index is
+# held at its 2025-26 level: journeys are assumed to stay flat.
+#
+# Dataset projection grows rail_usage by this index per household
+# (gov.economic_assumptions.yoy_growth.orr.per_capita.rail_passenger_journeys),
+# as household weights already grow by ons.population (#1862).
 values:
-  # Base year (COVID-affected)
-  2020-01-01: 1.000
-  # 2021: Recovery year
-  2021-01-01: 0.400
-  # 2022: Continued recovery
-  2022-01-01: 0.700
-  # 2023: Near full recovery
-  2023-01-01: 0.900
-  # 2024: Full recovery plus growth
-  2024-01-01: 0.950
-  # 2025: +1.9% growth (ORR/Steer projected trend)
-  2025-01-01: 0.968
-  # 2026: +1.9% growth
-  2026-01-01: 0.986
-  # 2027: +1.9% growth
-  2027-01-01: 1.005
-  # 2028: +1.9% growth
-  2028-01-01: 1.024
-  # 2029: +1.9% growth
-  2029-01-01: 1.044
+  2020-01-01: 1.0  # 387.885m
+  2021-01-01: 2.5525  # 990.062 / 387.885
+  2022-01-01: 3.5700  # 1,384.752 / 387.885
+  2023-01-01: 4.1559  # 1,612.028 / 387.885
+  2024-01-01: 4.4567  # 1,728.704 / 387.885
+  2025-01-01: 4.7165  # 1,829.468 / 387.885
+  2026-01-01: 4.7165  # held flat (assumption)
 metadata:
   unit: /1
   label: Rail ridership index
   reference:
-    - title: ORR Passenger Rail Usage Statistics
-      href: https://dataportal.orr.gov.uk/statistics/usage/passenger-rail-usage/
-    - title: Steer - UK Rail Demand Forecasting
-      href: https://www.steergroup.com/
+    - title: ORR, Table 1220 Passenger journeys, Great Britain, January 1872 to March 2026 (18 June 2026)
+      href: https://dataportal.orr.gov.uk/statistics/usage/passenger-rail-usage/table-1220-passenger-journeys/

policyengine_uk/parameters/gov/economic_assumptions/yoy_growth.yaml

@@ -1582,6 +1582,80 @@ finance_ni:
         - title: ONS, Country and regional public sector finances methodology guide, sections 29 and 30
           href: https://www.ons.gov.uk/economy/governmentpublicsectorandtaxes/publicsectorfinance/methodologies/countryandregionalpublicsectorfinancesmethodologyguide
 
+orr:
+  per_capita:
+    rail_passenger_journeys:
+      description: Growth in rail passenger journeys per person.
+      # Growth in gov.dft.rail.ridership_index (ORR Table 1220 passenger
+      # journeys, Great Britain) net of ons.population growth:
+      # (index_t / index_t-1) / (1 + population_t) - 1, rounded to 4
+      # decimals. Household weights grow by ons.population, so growing each
+      # household's rail_usage by this keeps the weighted total on ORR's
+      # journeys (#1862). The financial year starting in April of year t is
+      # stored at t-01-01. From 2026 journeys are held flat (no official
+      # forecast), so this is minus population growth, for every year
+      # ons.population has a value. Re-derive it whenever either series
+      # changes (policyengine_uk/tests/test_rail_usage_uprating.py checks).
+      values:
+        2021-01-01: 1.5433  # 2.5525 / 1.0036 - 1
+        2022-01-01: 0.3851  # (3.5700 / 2.5525) / 1.0098 - 1
+        2023-01-01: 0.1490  # (4.1559 / 3.5700) / 1.0132 - 1
+        2024-01-01: 0.0607  # (4.4567 / 4.1559) / 1.0110 - 1
+        2025-01-01: 0.0551  # (4.7165 / 4.4567) / 1.0030 - 1
+        2026-01-01: -0.0035  # 1 / 1.0035 - 1
+        2027-01-01: -0.0029  # 1 / 1.0029 - 1
+        2028-01-01: -0.0032  # 1 / 1.0032 - 1
+        2029-01-01: -0.0034  # 1 / 1.0034 - 1
+        2030-01-01: -0.0036  # 1 / 1.0036 - 1
+        2031-01-01: -0.0024  # 1 / 1.0024 - 1
+        2032-01-01: -0.0022  # 1 / 1.0022 - 1
+        2033-01-01: -0.0021  # 1 / 1.0021 - 1
+        2034-01-01: -0.0019  # 1 / 1.0019 - 1
+        2035-01-01: -0.0018  # 1 / 1.0018 - 1
+        2036-01-01: -0.0017  # 1 / 1.0017 - 1
+        2037-01-01: -0.0016  # 1 / 1.0016 - 1
+        2038-01-01: -0.0015  # 1 / 1.0015 - 1
+        2039-01-01: -0.0015  # 1 / 1.0015 - 1
+        2040-01-01: -0.0014  # 1 / 1.0014 - 1
+        2041-01-01: -0.0014  # 1 / 1.0014 - 1
+        2042-01-01: -0.0013  # 1 / 1.0013 - 1
+        2043-01-01: -0.0013  # 1 / 1.0013 - 1
+        2044-01-01: -0.0012  # 1 / 1.0012 - 1
+        2045-01-01: -0.0011  # 1 / 1.0011 - 1
+        2046-01-01: -0.0011  # 1 / 1.0011 - 1
+        2047-01-01: -0.0010  # 1 / 1.0010 - 1
+        2048-01-01: -0.0009  # 1 / 1.0009 - 1
+        2049-01-01: -0.0008  # 1 / 1.0008 - 1
+        2050-01-01: -0.0006  # 1 / 1.0006 - 1
+        2051-01-01: -0.0005  # 1 / 1.0005 - 1
+        2052-01-01: -0.0003  # 1 / 1.0003 - 1
+        2053-01-01: -0.0002  # 1 / 1.0002 - 1
+        2054-01-01: 0.0000  # 1 / 1.0000 - 1
+        2055-01-01: 0.0001  # 1 / 0.9999 - 1
+        2056-01-01: 0.0003  # 1 / 0.9997 - 1
+        2057-01-01: 0.0004  # 1 / 0.9996 - 1
+        2058-01-01: 0.0005  # 1 / 0.9995 - 1
+        2059-01-01: 0.0006  # 1 / 0.9994 - 1
+        2060-01-01: 0.0007  # 1 / 0.9993 - 1
+        2061-01-01: 0.0008  # 1 / 0.9992 - 1
+        2062-01-01: 0.0008  # 1 / 0.9992 - 1
+        2063-01-01: 0.0008  # 1 / 0.9992 - 1
+        2064-01-01: 0.0009  # 1 / 0.9991 - 1
+        2065-01-01: 0.0009  # 1 / 0.9991 - 1
+        2066-01-01: 0.0009  # 1 / 0.9991 - 1
+        2067-01-01: 0.0009  # 1 / 0.9991 - 1
+        2068-01-01: 0.0009  # 1 / 0.9991 - 1
+        2069-01-01: 0.0009  # 1 / 0.9991 - 1
+        2070-01-01: 0.0009  # 1 / 0.9991 - 1
+        2071-01-01: 0.0010  # 1 / 0.9990 - 1
+        2072-01-01: 0.0010  # 1 / 0.9990 - 1
+        2073-01-01: 0.0010  # 1 / 0.9990 - 1
+      metadata:
+        unit: /1
+        label: Rail passenger journeys growth per person
+        reference:
+          - title: ORR, Table 1220 Passenger journeys, Great Britain, January 1872 to March 2026 (18 June 2026)
+            href: https://dataportal.orr.gov.uk/statistics/usage/passenger-rail-usage/table-1220-passenger-journeys/
 carbon_intensity:
   description: Carbon intensity year-over-year growth (decline).
   values:

policyengine_uk/tests/code_health/test_uprating_declarations.py

@@ -24,11 +24,6 @@
 # Variables that declare an index but are deliberately not in the YAML, with
 # the reason. Anything else failing this test is a genuine gap.
 KNOWN_UNUPRATED = {
-    # gov.dft.rail.ridership_index is a CUMULATIVE index (2020 = 1.0), not a
-    # year-on-year growth rate, so it cannot go in a yoy_growth list without
-    # compounding wrongly: doing so uprates rail_usage by ~684% over 2024-27.
-    # Uprating this needs a yoy series, not a YAML entry (#1862).
-    "rail_usage",
     # Reported counterpart of employee_pension_contributions, which IS
     # uprated; the reported variable feeds imputation rather than results.
     "employee_pension_contributions_reported",

policyengine_uk/tests/test_projection_horizon.py

@@ -216,14 +216,12 @@ def test_unmapped_inputs_preserve_legacy_years_and_grow_after_2030(projected):
     """Previously unmapped columns keep their old values through 2030, then
     grow by the same cumulative-index ratio Core used beyond that year.
     Columns uprating_indices.yaml now lists (the debts, ISAs, lodger and
-    boarder rent, gross receipts, salary sacrifice and service charges), and
-    water bills, which grow by country, grow every year instead, so they are
-    no longer fallbacks.
+    boarder rent, gross receipts, salary sacrifice, service charges and rail
+    usage), and water bills, which grow by country, grow every year instead,
+    so they are no longer fallbacks.
     """
-    assert {
-        "rail_usage",
-        "employee_pension_contributions_reported",
-    } <= set(FALLBACK_INDEX_OF)
+    assert "employee_pension_contributions_reported" in FALLBACK_INDEX_OF
+    assert "rail_usage" not in FALLBACK_INDEX_OF
     for year in range(BASE_YEAR + 1, PROJECTION_END_YEAR + 1):
         for variable, index_name in FALLBACK_INDEX_OF.items():
             previous = _table(projected[year - 1], variable)[variable].to_numpy()

policyengine_uk/tests/test_rail_usage_uprating.py

@@ -0,0 +1,129 @@
+"""Rail usage is projected by ORR passenger journeys per person (#1862).
+
+ORR Table 1220 (published 18 June 2026) gives passenger journeys on Great
+Britain's railways in millions, by financial year. gov.dft.rail.ridership_index
+divides them by the 2020-21 figure. No official body forecasts journeys, so
+from 2026-27 they are held flat.
+
+Household weights grow by ons.population, so each household's rail_usage
+grows by journeys growth net of population growth. Growing it by journeys
+growth itself would count population growth twice.
+"""
+
+import numpy as np
+import pandas as pd
+import pytest
+
+from policyengine_uk import parameters
+from policyengine_uk.data import UKSingleYearDataset
+from policyengine_uk.data.economic_assumptions import (
+    PROJECTION_END_YEAR,
+    extend_single_year_dataset,
+)
+
+RIDERSHIP = parameters.gov.dft.rail.ridership_index
+POPULATION = parameters.gov.economic_assumptions.yoy_growth.ons.population
+PER_PERSON = (
+    "gov.economic_assumptions.yoy_growth.orr.per_capita.rail_passenger_journeys"
+)
+
+# ORR Table 1220, passenger journeys (million), by the year each financial
+# year starts in (2020 is April 2020 to March 2021).
+JOURNEYS = {
+    2020: 387.885467,
+    2021: 990.062033,
+    2022: 1_384.752483,
+    2023: 1_612.028410,
+    2024: 1_728.704107,
+    2025: 1_829.468293,
+}
+
+
+@pytest.mark.parametrize("year", sorted(JOURNEYS))
+def test_ridership_index_is_orr_journeys_on_a_2020_21_base(year):
+    # 2021-22: 990.062 / 387.885 = 2.5525; 2024-25: 1,728.704 / 387.885 =
+    # 4.4567. Before #1862 these were 0.400 and 0.950.
+    expected = JOURNEYS[year] / JOURNEYS[2020]
+    assert RIDERSHIP(f"{year}-01-01") == pytest.approx(expected, abs=5e-5)
+
+
+@pytest.mark.parametrize("year", range(2026, 2031))
+def test_journeys_are_held_flat_after_2025_26(year):
+    assert RIDERSHIP(f"{year}-01-01") == RIDERSHIP("2025-01-01")
+
+
+# Every year ons.population has a value for from 2021, so the offset runs as
+# far as the population series does and never stops while population growth
+# carries on changing.
+POPULATION_YEARS = [
+    int(value.instant_str[:4])
+    for value in POPULATION.values_list
+    if int(value.instant_str[:4]) >= 2021
+]
+
+
+@pytest.mark.parametrize("year", POPULATION_YEARS)
+def test_per_person_growth_is_journeys_growth_net_of_population_growth(year):
+    # 2025-26: (1,829.468 / 1,728.704) / 1.0030 - 1 = 0.0551. Fails if either
+    # input series changes without re-deriving this one.
+    instant, previous = f"{year}-01-01", f"{year - 1}-01-01"
+    expected = RIDERSHIP(instant) / RIDERSHIP(previous) / (1 + POPULATION(instant)) - 1
+    assert parameters.get_child(PER_PERSON)(instant) == pytest.approx(
+        expected, abs=5e-5
+    )
+
+
+def _dataset(rail_usage):
+    ids = np.arange(len(rail_usage))
+    return UKSingleYearDataset(
+        person=pd.DataFrame(
+            {
+                "person_id": ids,
+                "person_benunit_id": ids,
+                "person_household_id": ids,
+                "age": 40,
+            }
+        ),
+        benunit=pd.DataFrame({"benunit_id": ids}),
+        household=pd.DataFrame(
+            {
+                "household_id": ids,
+                "household_weight": 1.0,
+                "region": "LONDON",
+                "tenure_type": "OWNED_OUTRIGHT",
+                "rent": 0.0,
+                "council_tax": 0.0,
+                "rail_usage": rail_usage,
+            }
+        ),
+        fiscal_year=2024,
+    )
+
+
+def _total(dataset):
+    household = dataset.household
+    return (household["rail_usage"] * household["household_weight"]).sum()
+
+
+def test_projected_rail_usage_tracks_orr_journeys():
+    # Each household's usage grows by 1.0583 / 1.0030 = 1.0551 into 2025-26
+    # and the weights by 1.0030, so the weighted total grows by ORR's 5.83%.
+    # From 2026-27 the total stays flat in every projected year, past the end
+    # of the OBR forecast too. Before #1862 rail_usage was not projected at
+    # all.
+    projected = extend_single_year_dataset(
+        _dataset([100.0, 40.0]), parameters, end_year=PROJECTION_END_YEAR
+    )
+    datasets = projected.datasets
+    assert datasets[2025].household["rail_usage"].tolist() == pytest.approx(
+        [105.51, 42.204], rel=1e-4
+    )
+    assert _total(datasets[2025]) / _total(datasets[2024]) == pytest.approx(
+        JOURNEYS[2025] / JOURNEYS[2024], abs=1e-4
+    )
+    # Four-decimal rounding of the two growth series moves the total by at
+    # most 0.005% a year.
+    for year in range(2026, PROJECTION_END_YEAR + 1):
+        assert _total(datasets[year]) == pytest.approx(
+            _total(datasets[2025]), rel=5e-5 * (year - 2025)
+        ), year

policyengine_uk/tests/test_uprating_inputs_reach_data.py

@@ -46,6 +46,9 @@
 from policyengine_uk.utils.scenario import Scenario
 
 AVERAGE_EARNINGS = "gov.economic_assumptions.yoy_growth.obr.average_earnings"
+# The index employee_pension_contributions_reported declares. The YAML does
+# not list that variable, so only the post-2030 fallback reads its index.
+UNMAPPED_INDEX = "gov.economic_assumptions.indices.obr.average_earnings"
 PERSONAL_ALLOWANCE = "gov.hmrc.income_tax.allowances.personal_allowance.amount"
 HF_URL = "hf://policyengine/policyengine-uk-data-private/mock_frs_2025_26.h5"
 GCS_URL = "gs://policyengine-uk-data-private/mock_frs_2025_26.h5"
@@ -394,7 +397,7 @@ def recording_node_at(self, instant):
 
     covered = uprating_input_paths()
     # The unmapped indices, read past the legacy horizon, are in the set.
-    assert "gov.dft.rail.ridership_index" in read
+    assert UNMAPPED_INDEX in read
     for name in read:
         assert any(name == path or name.startswith(f"{path}.") for path in covered), (
             name
@@ -403,7 +406,7 @@ def recording_node_at(self, instant):
 
 def test_the_read_set_names_inputs_outside_the_growth_rates():
     paths = uprating_input_paths()
-    assert "gov.dft.rail.ridership_index" in paths
+    assert UNMAPPED_INDEX in paths
     assert economic_assumptions.CGT_INDEX_PATH in paths
     assert AVERAGE_EARNINGS in paths
 

policyengine_uk/variables/gov/dft/rail_usage.py

@@ -7,10 +7,13 @@ class rail_usage(Variable):
         "Rail usage quantity in base year (2020) price terms. "
         "Should be provided by policyengine-uk-data, derived from "
         "rail_subsidy_spending / fare_index at survey year. "
-        "Uprated by rail ridership growth trends (~1.9% annually)."
+        "Projected by ORR passenger journeys per person, held flat from "
+        "2026-27."
     )
     entity = Household
     definition_period = YEAR
     value_type = float
     unit = GBP
-    uprating = "gov.dft.rail.ridership_index"
+    # Per person, because household weights already grow by population; the
+    # growth is that of gov.dft.rail.ridership_index (#1862).
+    uprating = "gov.economic_assumptions.indices.orr.per_capita.rail_passenger_journeys"