rail_subsidy_spending should be uprated by regulated fare increases

[MaxGhenis]

Summary

The rail_subsidy_spending variable is an input from the dataset that grows very slowly (~0.4% per year), likely only from household weight changes. In reality, rail fares (and thus the value of rail subsidies to households) increase each year following the regulated fares formula (RPI + 1%).

Current behaviour

from policyengine_uk import Microsimulation

sim = Microsimulation()

for year in [2024, 2025, 2026, 2027, 2028, 2029]:
    val = sim.calculate('rail_subsidy_spending', year)
    print(f'{year}: £{val.sum()/1e9:.2f}bn')

Output:

2024: £10.70bn
2025: £10.78bn  (+0.7%)
2026: £10.82bn  (+0.4%)
2027: £10.86bn  (+0.4%)
2028: £10.90bn  (+0.4%)
2029: £10.95bn  (+0.4%)

Expected behaviour

Rail subsidy spending should grow with regulated fare increases:

• 2026: +5.8% (July 2025 RPI 4.8% + 1%)
• 2027: +4.2% (projected)
• 2028: +3.9% (projected)
• 2029: +3.9% (projected)

This matters for modelling rail fare freeze policies - the baseline should show fares increasing, so the reform (freeze) produces savings relative to that baseline.

Proposed solution

Add uprating to rail_subsidy_spending based on regulated fare increases. This could be:

1. A new parameter gov.dft.rail.regulated_fare_increase with RPI+1% values by year
2. Apply this as uprating to rail_subsidy_spending in the variable definition

References

• GOV.UK Rail Fares Freeze - states 5.8% counterfactual increase for 2026
• Regulated fares formula: July RPI + 1%

[MaxGhenis]

Investigation findings

The uprating doesn't work with Microsimulation because the FRS dataset provides the same rail_subsidy_spending value for all years (the raw values are identical, e.g., £2035.53 for household 93 in 2024, 2025, and 2026).

The uprating DOES work with simple Simulations:

from policyengine_uk import Simulation

sim = Simulation(
    situation={
        'people': {'person1': {'age': {2025: 30}}},
        'households': {
            'household1': {
                'members': ['person1'],
                'rail_subsidy_spending': {2024: 1000}
            }
        }
    }
)

# 2024: £1000.00
# 2025: £1044.64 (+4.5%)
# 2026: £1105.58 (+5.8%)
# 2027: £1151.93 (+4.2%)

Root cause: The multi-year dataset in policyengine-uk-data copies the same value across all years instead of letting the uprating system extrapolate.

Solution options:

1. Fix in policyengine-uk-data: Only store base year values and let uprating handle extrapolation
2. Fix in policyengine-uk-data: Apply rail fare index uprating when generating the multi-year dataset
3. Document that uprating parameters only work with Simulation, not Microsimulation with pre-filled data

Filed related issue in policyengine-uk-data: [TBD]