Set Marriage Allowance take-up rate to HMRC outturn (#368)

[vahid-ahmadi]

Summary

• Closes Set Marriage Allowance take-up rate to HMRC outturn (~50%) #368.
• Update policyengine_uk_data/parameters/take_up/marriage_allowance.yaml from 1.0 to 0.5, matching HMRC outturn (~2.1m claimants of ~4.2m eligible couples).
• Adds House of Commons Library reference and a description note that the file is read by datasets/frs.py to populate would_claim_marriage_allowance stochastically.

Why this matters

This file is the load-bearing one for microsimulation. With take-up at 1.0, the model overstates Marriage Allowance cost by ~£500m/year (model implies ~£1.06bn vs HMRC outturn ~£580m). At 0.5 it lines up with published outturn.

Companion change in policyengine-uk: PolicyEngine/policyengine-uk#1636 — the documentation parameter on the model side. Source issue: PolicyEngine/policyengine-uk#623.

Test plan

• FRS build run to confirm would_claim_marriage_allowance rate halves.
• Sanity check Marriage Allowance aggregate microsim cost against £580m HMRC outturn.

🤖 Generated with Claude Code

[vahid-ahmadi]

Review note: I checked this against the corresponding policyengine-uk change and the dataset build path in policyengine_uk_data/datasets/frs.py. I did not find a blocker here. This is the load-bearing parameter for baseline microsimulation because load_take_up_rate("marriage_allowance", year) is what populates would_claim_marriage_allowance during FRS build, so changing this file to 0.5 is the part that should move baseline outputs toward HMRC outturn. The companion policyengine-uk PR should stay in sync, but this PR itself looks internally consistent.

[MaxGhenis]

Verified the 50% figure and £580m outturn against HoC SN00870 and HMRC's December 2024 Non-structural tax relief statistics. Downstream path via frs.py → would_claim_marriage_allowance → marriage_allowance formula is clean (no calibration double-counting; calibration targets don't include MA aggregate). Pre-2015 effective date (2000-01-01) is harmless — the marriage_allowance amount parameters in policyengine-uk don't exist pre-2015 so the formula returns 0 regardless of take-up. Optional cleanup follow-up: move effective date to 2015-04-06.