[PavelMakarchuk]

Summary

Adds a UK blog post that explains the Budget 2026 Pension Tax Allowance (PTA) step by step. It walks through the rules, worked examples and household charts behind the PTA figures that the Autumn Budget dashboard publishes (£94, £182 and £250 full allowances), then the population results.

Depends on #1462 (the line chart type and bar chart valueLabels), which this PR is based on, so the diff here is the post only: the article, its posts.json entry and the cover.

Structure

1. What the PTA does: eligibility, value (20% of the full new State Pension above the Personal Allowance, £94 in 2027-28; Scottish taxpayers at 20%), a non-refundable tax reduction, the income test, the £1-in-£3 taper, and the tax-year table (£94/£182/£250, ending at £13,319/£14,025/£14,567).
2. Household impacts: the explainer's five pensioners; net income and other benefits for eight households (Pension Credit, Housing Benefit, Council Tax Reduction, Universal Credit for mixed-age couples, the TV licence for over-75s); the PTA in each tax year; income tax; net income; effective marginal tax rates.
3. Population impacts: budgetary impact beside Table 6.1 (−£42m, −£113m, −£193m against −£45m, −£90m, −£150m), who gains (528,000 households in 2027-28, 1.07 million in 2029-30), decile charts, poverty and inequality.
4. Assumptions and limitations, and Methodology.

Method

• Household calculations: pe.uk.calculate_household on published policyengine[uk]==6.11.1 (policyengine-uk 2.151.0, policyengine-core 3.33.3), on the OBR's October 2026 forecast as the dashboard's with_forecast applies it (uk-autumn-budget-dashboard-2026 at 64042657), with the PTA added to that scenario.
• The PTA reform: reduces income_tax, capped at the tax due, and also the tax Universal Credit deducts from earnings (uc_income_tax_on_earnings), which policyengine-uk computes separately (see the comment on uk-autumn-budget-dashboard-2026#290). Rules and values match budget_measures/pension_tax_allowance.py in Create Donate page #290.
• Pension Credit 2027-28: the household examples use £247.30 a week (the 2026-27 rate uprated by 3.9% earnings growth); the population run uses the model's £245.38, as the dashboard does.
• Population run: pe.uk.managed_microsimulation, as the dashboard's build_microsimulation does, on the certified Microcosm UK dataset the 6.11.1 bundle pins (microcosm-uk-2024-25-national-20261025T091414Z-a6299504). Static, no behavioural response.
• Charts: breakpoints of the piecewise-linear schedules plus a point every £100; they agree with PolicyEngine UK to the penny at every £2 of income, and the effective marginal rates to within 0.04 percentage points.
• Writing: whole pounds outside the two worked tables, as agreed for the Budget posts.

Checks

• Post, chart and content tests pass locally.
• Checked on a local dev server.

To review

• Post date: 2026-10-28 18:00.
• The cover: an original vector illustration (no licence needed).

🤖 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 4:37pm UTC
policyengine-website	Ready	Preview	Oct 28, 2026 4:37pm UTC

[MaxGhenis]

@PavelMakarchuk Max wants the Budget posts merged today so the dashboard's Policies tab can link them. Is it OK to merge this at 26feaea, after #1462?

Nothing from my review is open at 26feaea. Two checks remain. First, I'm running the PTA microsimulation on #290 at 9b98370e, the head the final regeneration runs from, to compare with the post's −£42m, −£113m and −£193m and its counts. Second, once #1462 merges and this PR retargets to main, CI has to pass again. If both hold, I'll mark it ready, approve and squash at 26feaea.

🤖 Generated with Claude Code

[MaxGhenis]

@PavelMakarchuk: I pushed bf81eaa on top of your 45a700b. It puts the population section back on the dashboard's basis, as the coordinator ruled under Max's "fix figures".

Why. The 45a700b rerun used the earnings-uprated Pension Credit guarantee (£247/£255/£260). The Autumn Budget dashboard (#290) uses PolicyEngine UK's own guarantee (£245 in 2027-28). Its PTA run gives a total cost of −£30m, −£83m and −£144m, the figures this post had at 26feaea. The post's population figures need to match the dashboard. Your household examples keep the earnings-uprated £247 guarantee, as they did before.

What changed (each back to its 26feaea value):

• Budgetary table, total cost: 2028-29 from −£82m to −£83m; 2029-30 from −£143m to −£144m. The income tax column (−£42m/−£113m/−£193m) is unchanged.
• Who gains: households gaining in 2027-28 from 529,000 to 528,000, and in 2028-29 from 855,000 to 860,000. Average gains in 2028-29 and 2029-30 go from £97 and £135 to £96 and £134.
• Losers: from 7,400 and 6,600 households (with the Housing Benefit, Council Tax Reduction and TV licence split) back to about 2,700 in 2028-29 and 2,800 in 2029-30, each losing a free TV licence.
• Both decile charts: the 26feaea values.
• Appendix, population run: back to "its own Pension Credit guarantee (£245 a week in 2027-28), as the dashboard does", with the £12,760 to £12,807 band.
• Comparison with the OBR: the sentence on the coincidence with PolicyEngine's total cost is back. With −£83m and −£144m it holds again.
• Assumptions: back to "For the household examples, we apply 3.9%... £247". I kept your Social Security Administration Act 1992, section 150A citation.

Winter Fuel Payment note: removed. It came with the earnings-basis run, and it also misstates the model. In winter_fuel_payment_eligible (PolicyEngine UK 2.151.0), someone who isn't on a relevant benefit keeps the payment if anyone of pension age in their household has their own total_income at or below £35,000 (England and Wales only). That is not a test on household income.

Please hold further pushes to #1457 until it's merged. The coordinator is merging #1462 and then this PR so the post is live for Q&A. If something needs changing, comment here and I'll fold it in.

🤖 Generated with Claude Code

[MaxGhenis]

@PavelMakarchuk: one more commit, 38cdd4a, and c9a4f38 for the losers line. It puts the remaining population figures on the dashboard's own run, so every number in the population section can be traced to it.

Sources

• Revision 20 artifact state_pension_income_tax_protection/db74e63711506e84:
  • winners_losers.csv gives the average-change decile chart.
  • distributional_impact.csv gives the share-of-income decile chart.
  • metrics.csv gives poverty and the Gini.
• A run of Create Donate page #290 at 9b98370e on the dashboard's builders (build_microsimulation) gives everything else. Its gov_balance change equals the artifact's budgetary impact to 12 digits, so it is the same simulation.

Changes

Item	Was	Now	Run value
Income tax, 2028-29	−£113m	−£112m	−£112.49m
Households gaining, 2028-29 / 2029-30	860,000 / 1.07m	857,000 / 1.09m	856,817 / 1,090,964
Their average gain	£96 / £134	£97 / £132	£96.71 / £131.85
Households losing	none / 2,700 / 2,800	fewer than 10,000 in each year (c9a4f38)	small sample, so no precise estimate
Decile charts	earlier run	artifact values	e.g. 2nd decile, 2027-28: £1.54 → £1.52
Gini	"less than 0.0001"	"less than 0.0001 in 2027-28 and about 0.0001 in 2029-30"	−0.000027 / −0.000125

Notes on the changes

• Losers. The old line said every loser loses a free TV licence. In the run, the TV licence change is +£0.49m in 2028-29 and +£0.51m in 2029-30. That is about 2,700 households, not all the losers, so the line now says "some of them". The sample is too small for a precise count of losers, so c9a4f38 gives them as fewer than 10,000 a year.
• Top deciles. The 2029-30 sentence on the top three deciles now says "about £1 or less", because the 8th decile is £1.16.
• Thresholds. A household gains or loses if its household_net_income changes by more than half a penny. A person pays less tax if their income_tax falls by more than half a penny.

Unchanged, and they match:

• total cost: −£30m, −£83m and −£144m;
• income tax: −£42m and −£193m;
• people paying less tax: 668,000, 1.06m and 1.36m;
• 528,000 households gaining in 2027-28, at £56;
• poverty: −0.01pp and −0.02pp.

All six chart blocks parse. The PR is still a draft, so it needs marking ready before the coordinator approves it.

🤖 Generated with Claude Code

[MaxGhenis]

Approving at c9a4f38 on Max's go, so Pavel can walk through the post in today's webinar Q&A. It's the first of the full explainers for each Budget measure.

• Review. The PTA review seat reviewed the post and found no must-fix items. I reviewed the last three commits' diffs myself.
• Figures match the dashboard.
  • Total cost is −£30m, −£83m and −£144m. That is revision 20's artifact for state_pension_income_tax_protection, and the seat's own run at #290's head reproduces the dashboard's budgetary impact to 12 digits.
  • The decile charts, poverty and Gini come from the same artifact.
  • The household counts come from that same simulation.
• Population basis. The section uses the model's Pension Credit guarantee, as the dashboard does. The household examples keep the earnings-uprated guarantee.
• Losers. The sentence gives no figures from small sample cells.
• Diff. Only the post, its cover image and one posts.json entry. The chart support merged in #1462.

🤖 Generated with Claude Code