[MaxGhenis]

Hold: don't merge. This changes public text on the live briefing (#1469). It waits for the Budget coordinator or Max to say go.

The change

One sentence in the briefing's Pension Tax Allowance comparison (L200), from Pavel's live review, part 5, item 3:

• Was: "We estimate that about 670,000 people gain in 2027-28, against the Treasury's 500,000."
• Now: "We estimate that about 670,000 people above State Pension age pay less income tax in 2027-28, against the Treasury's 500,000."

The figure doesn't change. Only the words that say what it counts change.

Why

• The 670,000 is people_directly_affected in the Pension Tax Allowance artifact: public/data/artifacts/state_pension_income_tax_protection/1bfeed3c27a3c9b9/age_impact.csv, 2027, all row, 665,907.
  • calculators.py counts a person when their own income tax, National Insurance, student loan repayment or employment income changes by £1 a year or more (DIRECT_EFFECT_VARIABLES, DIRECT_EFFECT_MIN_GBP).
  • The measure's modifier (budget_measures/pension_tax_allowance.py) replaces only income_tax and Universal Credit's two income-tax-on-earnings slices. It sets the reduction to zero unless is_SP_age, and it can only lower income tax (max(0, current - used)).
  • In a local policyengine-uk checkout (not pinned to 2.151.0), none of the other three variables reads income_tax directly: NI adds its contribution components, student loan repayment reads adjusted_net_income, and employment income sums its components.
  • Pavel's review and the Pension Tax Allowance post read the count the same way: people at State Pension age whose income tax falls. The 65+ row holds all 665,907.
• Everywhere else, the post's "gain" means people who live in households that gain (Table 5 and the measure paragraphs). On that measure about 990,000 people gain in 2027-28: the same all row's pct_better_off, 1.41%, of its people, 69.99 million. metrics.csv gives the same 1.41% as people_affected.
• The Treasury's 500,000 counts pensioners who pay less income tax, in HM Treasury's explainer as the Pension Tax Allowance post quotes it (L757). That post also says both counts cover "people over State Pension age whose income tax falls". The new wording compares like with like. Pavel's review (part 5, item 3) proposed the change.

Not changed

• The dashboard note "More pensioners gain in our data (670,000 vs 500,000)" (obrDeviationNotes.js in the dashboard repo). It belongs to the dashboard, so I'm passing it to the Budget coordinator to route to Max.
• Nothing else in the post. Pavel's other points are wording, or open items already in hand: Scroll wide research-post tables inside their own container #1472, dashboard Browser back button doesn't move backward through frames #303 and decision d2159. The reply on Autumn Budget 2026: post-Budget written briefing #1469 answers them point by point.

Checks

• git diff touches one line of uk-autumn-budget-2026-briefing.md.
• Independent review: GPT-6.1 Sol. The result goes in a comment below.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
policyengine-calculator-next	Ready	Preview	Oct 29, 2026 7:11am UTC
policyengine-website	Ready	Preview	Oct 29, 2026 7:11am UTC

[MaxGhenis]

Independent review: GPT-6.1 Sol, two rounds, at head 0a21ff4.

• r1 (20261029-031349-briefing-pavel-live-r1): REQUEST CHANGES. It questioned whether "above State Pension age" holds, and it found that this body cited a column in the wrong file. It found no other error in the post among Pavel's points.
• r2 (20261029-032127-briefing-pavel-live-r2): APPROVE WITH CHANGES. With the modifier's code in front of it (budget_measures/pension_tax_allowance.py gates the reduction on is_SP_age and changes only income tax), it withdrew the age finding and kept the sentence as it stands. Its two wording points, on the dependency evidence and on attribution, are now in the body above.

Sol checked:

• the article hash;
• that the diff changes one sentence;
• the 665,907 direct count and the roughly 988,000 people in gaining households;
• sentence case and the "exceeds" rule.

CI: 10 of 10 checks pass, and GitHub reports the PR MERGEABLE. Still held: this waits for the Budget coordinator's or Max's go.

🤖 Generated with Claude Code

[MaxGhenis]

Merged at 0a21ff4 with --admin --match-head-commit on Max's go (d2170, 29 Oct ~14:25Z, via the Budget coordinator). The only block was the required review, which no one can give on Max's own PR. Gates: gh pr checks exit 0 (10/10); MERGEABLE; no CHANGES_REQUESTED; GPT-6.1 Sol r2 approved the one changed sentence as it stands (its two points were on this PR body, now in it); the coordinator read the diff (one sentence, 670,000 unchanged, now counts people above State Pension age who pay less income tax).

🤖 Generated with Claude Code