juaristi22 · Oct 21, 2026 · Oct 20, 2026 · Oct 20, 2026 · Oct 20, 2026 · Oct 21, 2026
diff --git a/changelog.d/2542.fixed.md b/changelog.d/2542.fixed.md
@@ -0,0 +1,2 @@
+- Bus fares from journeys now follow DfT's published local bus fares index (BUS0415a) by area and fiscal year from the 2024-25 benchmark, and the OBR CPI after the last published quarter (March 2026), instead of holding the benchmark yield and adding a cap term against a fixed 2024 cap schedule. The 2025-26 yield per fare-paying boarding outside London rises 12.6% (the index) instead of 24.9%. Northern Ireland, which has no published index, follows CPI. Cap changes after the last published quarter (the £2 cap in 2027 and its end in 2028) enter the baseline through the cap term. The bus fare cap schedules now keep their calendar dates, so a cap change part-way through a year applies from its date.
+- Cap changes are priced from a three-band table of uncapped single fares (`gov.dft.bus.uncapped_single_fares`), fitted to DfT's evaluation of the £2 cap, the 2023-24 fare cap grant, the January 2025 fares index step and the funding announced for the £3 cap in 2025 and the £2 cap in 2027, and uprated with CPI, instead of one £2.748 fare for every single ticket. A £3 cap now binds, and a full year at £2 against no cap saves about £0.36bn a year at 2024-25 boardings instead of £0.70bn. A supplied `local_bus_uncapped_single_fare` replaces the table for that person.
diff --git a/changelog.d/2542.removed.md b/changelog.d/2542.removed.md
@@ -0,0 +1 @@
+- Removed `gov.dft.bus.reference_year` and `gov.dft.bus.uncapped_single_fare`. The benchmark year of the published bus fare yield is `gov.dft.bus.benchmark_year`; the cap term reads `gov.dft.bus.fares.cap` and `gov.dft.bus.reference_fares.cap` in the same year; and cap changes are priced from `gov.dft.bus.uncapped_single_fares`. A reform naming a removed parameter fails with directions.
diff --git a/docs/engineering/bus-fares.md b/docs/engineering/bus-fares.md
@@ -27,8 +27,9 @@ false and the person pays fares even if eligible for a statutory concession.
 For fare-paying passengers,
 `local_bus_single_fare_share` is the share of boardings bought as singles,
 bounded to [0, 1]. Its fallback is 0.70, the 2024 NTS other-local-bus estimate.
-An optional `local_bus_uncapped_single_fare` overrides the representative
-uncapped fare described below. Negative quantities and fares price at zero.
+An optional `local_bus_uncapped_single_fare` replaces the table of uncapped
+single fares described below for that person. Negative quantities and fares
+price at zero.
 
 On loading a UK `Simulation`, a supplied household `bus_fare_spending` moves
 to `bus_fare_spending_reported`. The stored total therefore cannot freeze the
@@ -66,6 +67,59 @@ residents because non-residents also travel there. Scotland and Northern
 Ireland use the English other-local-bus trip rate as a proxy, as Microcosm does.
 These translations include interchanges and differences between operator
 boardings and resident trips; they are not counts of legs for a known trip.
+`gov.dft.bus.benchmark_year` (2024) names the fiscal year these facts describe.
+
+## Fares by year
+
+In other years each area's yield moves with its published fares index:
+
+```
+yield(t) = benchmark_yield * index(t) / index(benchmark_year)
+```
+
+`gov.dft.bus.local_bus_fares_index` holds DfT's
+[BUS0415a](https://www.gov.uk/government/statistical-data-sets/bus04-costs-fares-and-revenue)
+quarter-end values (March 2005 = 100) for London, England outside London and
+Scotland, from DfT's June 2026 release (quarters to March 2026;
+[archived copy](https://web.archive.org/web/20260619124752/https://assets.publishing.service.gov.uk/media/6a310ac715f2a70fac7e6006/bus0415.ods)),
+the vintage Chronicle and Microcosm vendor. DfT's September 2026 release, now
+served at the same address, revises the last two quarters and adds June 2026;
+move to it together with Microcosm. Each value is dated at the start of the
+quarter it ends, and `fiscal_year_blend` day-weights the quarters over 6 April
+to 5 April, the model's year. FY2024–25 outside London is therefore
+(86 × 181.6 + 92 × 185.7 + 92 × 185.9 + 90 × 206.4 + 5 × 210.5) / 365 = 190.23,
+within 0.2% of the mean of its four quarter ends (189.9). After the last
+published quarter the index grows with the OBR consumer price index each
+January, as `private_rent_index` does. Northern Ireland has no published index,
+so its yield grows with CPI from the benchmark year. Wales stays unpriced.
+
+The CPI projection also raises the fares in the index that the £3 cap holds at
+£3, while the cap term below measures changes from a nominal £3. From January
+2027 a ticket held at a cap therefore costs about £3 times the CPI growth since
+March 2026 more than the cap allows. That overstates the outside-London yield
+by 0.4% in 2027–28 and 0.8% in 2028–29, about 0.3 points more each year after.
+A cap reform is measured against the same reference as the baseline, so its
+cost does not move.
+
+| Area | FY2024–25 | FY2025–26 | FY2026–27 | FY2027–28 |
+|---|---:|---:|---:|---:|
+| London | 196.10 | 196.10 | 197.13 | 201.08 |
+| England outside London | 190.23 | 214.18 | 219.09 | 223.48 |
+| Scotland | 210.63 | 229.94 | 235.90 | 240.63 |
+
+FY2025–26 and earlier are published; later years are CPI projections. The
+index records prices paid, so it already contains each cap in force while it
+is published, including the rise to £3 in January 2025 (outside London 185.9
+in December 2024, 206.4 in March 2025). The cap term below adds only what the
+index cannot contain.
+
+Each area follows its own index, so the England total moves with the
+receipts-weighted mix of the London and outside-London indices: +7.6% in
+FY2025–26. DfT's England index rose 6.4%. Its step from December 2024 to
+March 2025 was 5.6%, against 11.0% outside London and none in London, so it
+weights outside London at about one half, where receipts weight it at 61%.
+A receipts target moved by the England index therefore sits below the model's
+England total.
 
 ## Caps and the uncapped counterfactual
 
@@ -79,45 +133,114 @@ representing the end of the announced scheme. These dates follow the
 The cap does not affect the London series, Scottish or Northern Irish fares.
 England outside London uses a representative participating-service assumption;
 route-level participation and local lower caps require caller-supplied fares.
+The table of uncapped fares below is fitted to national figures and applied
+the same way in every English region outside London. Where a combined authority
+already runs its own lower single fare, a national cap reform therefore
+overstates the savings, and elsewhere it understates them. National totals
+hold, but regional and distributional splits of a cap reform lean the wrong way.
 
-The published yield already includes ticket discounts and the benchmark cap.
-For an English fare-paying person's other-local-bus boardings, the model uses:
+The published yield and index already include ticket discounts and the cap in
+force. `gov.dft.bus.reference_fares.cap` records that cap: the cap in force up
+to the last published quarter of the index (£3 from January 2025), held after
+it. For an English fare-paying person's other-local-bus boardings, the model
+uses:
 
 ```
-current_yield = max(0, benchmark_yield + single_share *
-                   (annual_min(uncapped_fare, current_cap) -
-                    annual_min(uncapped_fare, reference_cap)))
-spending = trips * boardings_per_trip * current_yield
+yield(t) = max(0, benchmark_yield * index(t) / index(benchmark_year)
+                  + single_share * sum over bands b of share_b *
+                    (annual_min(fare_b, fares.cap, t) -
+                     annual_min(fare_b, reference_fares.cap, t)))
+spending = trips * boardings_per_trip * yield(t)
 ```
 
+Both schedules are read in the same year. While the index is published they
+agree and the cap term is zero, so the baseline is the published path. After
+the last published quarter the term adds the scheduled changes: the £2 cap from
+January 2027 and its end in 2028. A reform of `fares.cap` adds its own
+difference from the reference. Leave `reference_fares` unchanged in reforms.
+When a newly published quarter is added to the index, extend
+`reference_fares.cap` with the cap then in force;
+`policyengine_uk/tests/test_bus_fare_parameters.py` fails until the two
+schedules agree up to the last published quarter.
+
 `annual_min` caps the fare in each date interval, then weights the intervals
 from 6 April to 5 April, assuming uniform journeys through the year. It does
-not cap a fare using the average cap. Both `fares` and `reference_fares` preserve
-calendar dates. `reference_year` is 2024 and `reference_fares` holds the benchmark
-schedule fixed when the policy cap changes. The source receipts use 1 April to
-31 March; this approximation treats them as the corresponding model year's
-benchmark. Annual Scenario changes apply to the full model fiscal year.
-
-The default uncapped fare is £2.748019953976. This is an aggregate assumption,
-not an observed fare distribution. The FY2024–25 cap grant is £515,693,216.60.
-With `F = B - C` outside London and single share `s = 0.70`, the assumption
-solves `s * (u - annual_min(u, reference_cap)) = grant / F`. Since `2 < u < 3`,
-the reference year has 270 days at £2 and 95 at `u`, giving:
+not cap a fare using the average cap. Both `fares` and `reference_fares` keep
+calendar dates (their index files propagate `preserve_calendar_dates` to the
+cap leaves), so a cap change part-way through a year applies from its date.
+The source receipts use 1 April to 31 March; this approximation treats them as
+the corresponding model year's benchmark. Annual Scenario changes apply to the
+full model fiscal year.
 
-```
-u = 2 + grant / (F * 0.70 * 270/365)
-```
+### Uncapped single fares
+
+What a cap saves depends on how uncapped single fares are spread, not on their
+average. `gov.dft.bus.uncapped_single_fares` holds three bands, each with its
+share of single tickets and its mean uncapped fare at FY2024–25 prices:
+
+| Band | Share | Mean fare |
+|---|---:|---:|
+| `LOW`, at or below £2 | 0.55 | £1.65 |
+| `MIDDLE`, above £2 and up to £3 | 0.30 | £2.21 |
+| `HIGH`, above £3 | 0.15 | £3.87 |
+
+The fares rise with the OBR CPI after FY2024–25 and the shares stay, so a fixed
+cap saves more as fares rise. The model caps each band's mean fare under both
+schedules and weights the bands by their shares, which must sum to one. With
+the bands at their means, the table is exact for caps of £2 and £3 at
+FY2024–25 prices and approximate for other caps; a cap below the low band's
+mean needs more bands.
+
+`scripts/fit_bus_cap_savings.py` fits the table by least squares to the facts
+below, each scaled by its tolerance. It searches shares in steps of one point
+and fares in steps of one penny, keeps each band's mean at least 10% above its
+lower edge, and sets the low band's fare so all singles average £2.15:
+
+| Fact | Source | Target | Table gives |
+|---|---|---:|---:|
+| A full year at £2, 2023–24 | Fare cap support, DfT BUS05bi | £278.95m ± 10% | £280.9m |
+| Part of the December 2024 to March 2025 index step outside London due to the £3 cap | BUS0415a: +11.0%; neighbouring quarters 0.1% to 2.3% | 10% ± 1 point of the yield | 9.8% |
+| The £3 cap in 2025 | ["over £150 million"](https://www.gov.uk/government/speeches/national-bus-fare-cap) | £150m ± 20% | £140.0m |
+| The £2 cap in 2027, against no cap | ["£400 million"](https://www.gov.uk/government/speeches/2-bus-fares-from-january-2027) | £400m ± 20% | £411.6m |
+| Mean pre-cap price of single tickets | [DfT evaluation](https://www.gov.uk/government/publications/evaluation-of-the-2-bus-fare-cap), Table 6 | £2.15 ± £0.10 | £2.15 |
+| Share of single tickets above £2 | Table 6: £0.32 / (£2.73 − £2.00) | 0.44 ± 0.05 | 0.45 |
+| Their mean pre-cap price | Table 6 | £2.73 ± £0.10 | £2.76 |
+| 2024–25: £2 for 270 days, then £3 | Fare cap support, DfT BUS05bi | £515.69m, not fitted | £268.5m |
+
+The £400 million is read as £2 against no cap: the £3 scheme is funded only to
+December 2026, and the £2 scheme replaces it. Read as £2 against a continued
+£3 cap, the table gives £247.3m, 38% below. The script prints both.
+
+The 2024–25 grant, for the benchmark year itself, is reported and not fitted;
+the table puts it at about half. Grant per fare-paying boarding rose 78% on
+2023–24, while the outside-London index rose 6.7%, receipts per fare-paying
+boarding rose 2.8%, and the cap rose only for the last quarter. The figure may
+therefore hold more than that year's forgone fares (timing or coverage of
+payments, not checked). The 2023–24 grant and the two announced figures all
+put a full year at £2 near £0.3bn, not £0.6bn. The single £2.748 fare this
+table replaced was calibrated to the 2024–25 grant.
+
+The costs use the 0.70 default single share and fare-paying boardings outside
+London (BUS01 journeys less BUS01c concessionary journeys: 1,281m in 2023–24
+and 1,331m in 2024–25, which also stands for 2025 and 2027), with fares moved
+to each year by CPI. Table 6 compares January to October 2022 with the same
+months of 2023 and is read at face value, for the table's shape. Deflated to
+2022 prices it would put the 2023–24 saving about a third above the grant.
+Possible reasons, not checked, are that the grant pays operators for revenue
+lost net of the journeys the cap added, and that Table 6 covers the larger
+participating operators only. The two announced figures are budgets; the £2
+cap's first ten months paid £210m against £245m planned. The table is a fitted
+central estimate, not an observed fare distribution.
+
+At FY2025–26 prices the table saves £0.39 per single ticket for a full year at
+£2 against no cap, and £0.15 at £3: about £0.36bn and £0.14bn a year at
+2024–25 boardings. The single £2.748 fare this replaced saved £0.75 at £2
+(about £0.70bn) and nothing at £3.
 
-Removing the cap therefore raises the benchmark yield from about £1.55 to
-£1.94 at the representative share, matching the grant-addback calculation
-in issue #1871. This assumes full grant pass-through and no demand response.
-Individual shares can differ, so a dataset aggregate need not recover the
-published grant. Caps at or above £2.75, including the £3 cap, have no effect
-relative to uncapped travel under the default £2.748 representative fare.
-The default has no fare distribution above these caps, so it cannot estimate
-the savings those caps give to passengers with higher fares. Supply
-person-specific `local_bus_uncapped_single_fare` values above a cap to model
-its effect on those passengers. Very large reductions can reach the zero-yield
+A dataset or calculator can supply `local_bus_uncapped_single_fare` for each
+person. Where it is supplied, that person's single tickets cost that one fare
+before the cap, and the table is not used. Single-year datasets extended by
+`Simulation` uprate it with CPI. Very large reductions can reach the zero-yield
 floor.
 
 ## Other reform controls and limits
@@ -127,8 +250,8 @@ spending. Both default to one; they are neutral reform controls rather than
 historical index series. Keep the denominator positive. `ridership_index`
 defaults to one and multiplies journey-priced spending, leaving supplied
 journey inputs intact. The model includes no demand elasticity or automatic
-ridership growth. Benchmark yields and trip translations stay nominally
-constant after 2024 unless changed explicitly. Dataset CPI-uprated household
+ridership growth. Trip translations stay constant after 2024 unless changed
+explicitly; yields move with the fares index. Dataset CPI-uprated household
 bus totals do not determine prices where journeys are available.
 
 A proportional scenario can set `gov.dft.bus.fare_index` to 0.9 for a 10%

diff --git a/policyengine_uk/data/uprating_indices.yaml b/policyengine_uk/data/uprating_indices.yaml
@@ -65,6 +65,7 @@ gov.economic_assumptions.yoy_growth.obr.consumer_price_index:
 - income_support_reported
 - jsa_contrib_reported
 - jsa_income_reported
+- local_bus_uncapped_single_fare
 - maintenance_expenses
 - maternity_allowance_reported
 - miscellaneous_consumption

diff --git a/policyengine_uk/parameters/gov/dft/bus/benchmark_year.yaml b/policyengine_uk/parameters/gov/dft/bus/benchmark_year.yaml
@@ -0,0 +1,15 @@
+description: Fiscal year of the published bus fare benchmark. yield_per_boarding and
+  boardings_per_trip describe this year (FY2024-25 receipts and boardings), and the yield in
+  another model year is the benchmark yield times the local bus fares index in that year over
+  its value in this one.
+values:
+  '0001-01-01': 2024
+metadata:
+  unit: year
+  period: year
+  label: bus fare benchmark year
+  reference:
+  - title: DfT bus statistics, BUS01 and BUS05i, 2024-25
+    href: https://www.gov.uk/government/statistical-data-sets/bus-statistics-data-tables
+  - title: Microcosm 954, published-fact extraction and pricing derivation
+    href: https://github.com/PolicyEngine/microcosm/pull/954
diff --git a/policyengine_uk/parameters/gov/dft/bus/fare_index.yaml b/policyengine_uk/parameters/gov/dft/bus/fare_index.yaml
@@ -1,5 +1,6 @@
 description: Proportional multiplier on bus fares; a neutral reform lever, not a measured
-  historical fare index.
+  historical fare index. The published DfT fares index that moves the baseline is
+  gov.dft.bus.local_bus_fares_index.
 values:
   '0001-01-01': 1
 metadata:

diff --git a/policyengine_uk/parameters/gov/dft/bus/fares/index.yaml b/policyengine_uk/parameters/gov/dft/bus/fares/index.yaml
@@ -1,2 +1,6 @@
 metadata:
+  # The formula day-weights the cap over 6 April to 5 April itself, so the
+  # leaves keep their calendar dates instead of fiscal-year annualisation.
+  # Leaves only see node metadata that is propagated.
   preserve_calendar_dates: true
+  propagate_metadata_to_children: true
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1,2 @@
		- Bus fares from journeys now follow DfT's published local bus fares index (BUS0415a) by area and fiscal year from the 2024-25 benchmark, and the OBR CPI after the last published quarter (March 2026), instead of holding the benchmark yield and adding a cap term against a fixed 2024 cap schedule. The 2025-26 yield per fare-paying boarding outside London rises 12.6% (the index) instead of 24.9%. Northern Ireland, which has no published index, follows CPI. Cap changes after the last published quarter (the £2 cap in 2027 and its end in 2028) enter the baseline through the cap term. The bus fare cap schedules now keep their calendar dates, so a cap change part-way through a year applies from its date.
		- Cap changes are priced from a three-band table of uncapped single fares (`gov.dft.bus.uncapped_single_fares`), fitted to DfT's evaluation of the £2 cap, the 2023-24 fare cap grant, the January 2025 fares index step and the funding announced for the £3 cap in 2025 and the £2 cap in 2027, and uprated with CPI, instead of one £2.748 fare for every single ticket. A £3 cap now binds, and a full year at £2 against no cap saves about £0.36bn a year at 2024-25 boardings instead of £0.70bn. A supplied `local_bus_uncapped_single_fare` replaces the table for that person.
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		- Removed `gov.dft.bus.reference_year` and `gov.dft.bus.uncapped_single_fare`. The benchmark year of the published bus fare yield is `gov.dft.bus.benchmark_year`; the cap term reads `gov.dft.bus.fares.cap` and `gov.dft.bus.reference_fares.cap` in the same year; and cap changes are priced from `gov.dft.bus.uncapped_single_fares`. A reform naming a removed parameter fails with directions.