Add ONS land value calibration targets

[vahid-ahmadi]

Summary

• Adds 4 aggregate land/property wealth calibration targets from the ONS National Balance Sheet 2025, so the weight optimiser can pull model estimates toward officially measured values
• The model currently estimates ~£10.8tn total land for 2026, but ONS measured £7.1tn for 2024 — without calibration targets the uprating indices overshoot
• Targets: household_land_value (£5.04tn), corporate_land_value (£2.06tn), land_value (£7.10tn), property_wealth (£7.10tn)

Changes

File	Change
targets/sources/ons_land_values.py	New target source: 4 targets derived from ONS 2024 measured total (£7.1tn), splitting household/corporate proportionally from 2020 breakdown
targets/compute/other.py	New compute_land_value() function
targets/compute/__init__.py	Export compute_land_value
targets/build_loss_matrix.py	Import + dispatch routing for the 4 land/property target names

Sources

• ONS National Balance Sheet 2025: https://www.ons.gov.uk/economy/nationalaccounts/uksectoraccounts/bulletins/nationalbalancesheet/2025
  • 2020 household land: £4.31tn, corporate land: £1.76tn, total: £6.07tn
  • 2024 directly measured total: £7.1tn
• The 2020 household/corporate split is scaled proportionally to match the 2024 measured total of £7.1tn
• The same £7.1tn target is used for 2024–2026 as no newer ONS measurement is available

Test plan

• Verified targets are discoverable via get_all_targets(year=2026) — returns all 4 with correct values
• Run full calibration pipeline to confirm targets are included in the loss matrix
• Compare pre/post calibration land value aggregates

🤖 Generated with Claude Code