[MaxGhenis]

Max found the official-comparison paragraph in the live Carer's Allowance taper post (#1464) confusing. This splits it into three paragraphs: the count, the cost and the official comparison. No figures change.

Before (one paragraph):

We find more carers whose allowance rises than the government estimates will benefit, about 82,000 against its 48,000 in 2030-31, and we model no change in hours worked. Our extra Carer's Allowance, £160 million in 2031-32, comes before any offset: Universal Credit takes back about half of it and income tax about a fifth, leaving a net cost of £24 million in 2030-31 and £48 million in 2031-32. Of the official £38 million and £79 million, … Before income tax, our figure is £39 million and £79 million, about a sixth more than that spending. …

After:

For 2030-31, DWP estimates that around 48,000 people gain a higher award, including around 3,000 in Wales, against our 82,000. Both counts compare awards with and without the taper and include people who become newly entitled, but DWP's also depends on assumptions about how carers respond, while ours assumes they work the same hours.

Of our £160 million of extra Carer's Allowance in 2031-32, lower Universal Credit offsets about half and higher income tax about a fifth, leaving a net cost of £48 million. In 2030-31, with half a year of the taper, the net cost comes to £24 million.

The government's costing, £38 million in 2030-31 and £79 million in 2031-32, measures something different from ours. It includes £5 million and £10 million of extra block grant for Scotland, which gained powers over Carer's Allowance in 2018. The other £33 million and £69 million cover the extra Carer's Allowance less the Universal Credit saving, and the costing has no entry for income tax. Against those £33 million and £69 million, our extra Carer's Allowance less lower Universal Credit comes to £39 million and £79 million, about a sixth more. Neither the Treasury nor the OBR splits the Carer's Allowance cost from the Universal Credit saving.

Other changes:

• Line 33. It names the Department for Work and Pensions (DWP), the source the data sources document gives for the 48,000 (p. 30). It also attaches the 3,000 in Wales to the 48,000, not to the newly entitled.
• Line 759. It drops "Lower Universal Credit and higher income tax offset part of the cost of the higher allowance", because the new cost paragraph says it with the shares.
• Next paragraph. "the Treasury's assumption" becomes "the government's assumption", and "our figures and the Treasury's" becomes "… the official ones", matching "the government's costing".

What changed and why:

• Static assumption. The old "we model no change in hours worked" is gone. The section's opening paragraph and the next paragraph both state it, and the count comparison now contrasts it with DWP's behavioural assumptions (p. 30: the estimates "are sensitive to assumptions about earnings, take-up and carer's behavioural responses").
• "£79 million". It meant both the official total and our figure before income tax. The comparison now leads with the £33m/£69m it compares against.
• Years. Each sentence of the cost paragraph covers one year.

Checks:

• An independent in-session Opus review (round 1 CHANGES, all applied or adjusted) re-checked every sentence against these sources:
  • OBR policy Table 1.1, measure 7 (−33.17/−68.73 welfare, −4.92/−10.16 Scottish BGA, −38.09/−78.90 total; no income tax line);
  • the data sources document, p. 30;
  • the released dashboard components (CA +78.45/+160.06, UC −39.21/−80.64, income tax +15.45/+31.84).
• The delta at b28ee86 got APPROVE.
• A grep of the section finds no passives and no "is/are above/below/around".

🤖 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 28, 2026 10:03pm UTC
policyengine-website	Ready	Preview	Oct 28, 2026 10:03pm UTC

[MaxGhenis]

Admin-merged at b28ee86 on Max's explicit go ("Admin-merge now"). The only block was the required approving review, which Max can't give on a PR under his own account.

• Independent review: an in-session Opus reviewer asked for changes in round 1; all were applied or adjusted. Its delta review approved b28ee86.
• CI: gh pr checks exit 0, MERGEABLE, no CHANGES_REQUESTED.
• Scope: wording only, in one article file. Every figure is unchanged and was checked against OBR policy Table 1.1 measure 7, the data sources document p. 30, and the released dashboard components.

Anthony, please flag anything you'd word differently and I'll follow up.