Skip to content

Add benefits-in-kind scoping deck for NEF - #119

Open
vahid-ahmadi wants to merge 11 commits into
mainfrom
vahid/nef-bik-scoping
Open

vahid-ahmadi wants to merge 11 commits into
mainfrom
vahid/nef-bik-scoping

Conversation

@vahid-ahmadi

@vahid-ahmadi vahid-ahmadi commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

Summary

A 16-slide deck at /slides/nef-bik-2026 for a scoping meeting with the New Economics Foundation on modelling employer-provided benefits in kind in PolicyEngine UK. Nothing has been agreed and nothing has been run for it: the deck has no PolicyEngine results.

Three parts, each with a section slide and a header tracker:

  1. Benefits in kind in the UK: what a benefit in kind is and how it is taxed; its size (HMRC, 2024-25 provisional, with the last split by type, 2016-17, in its own panel as shares); the rule changes to 2030-31
  2. What PolicyEngine covers: the pipeline with company cars carried through each stage, and a coverage map by benefit. Both carry status chips (Done, Partly done, Not yet) with a key. Company cars read Partly done on rules and on data (no OpRA, shared or pool cars; part-year use and employee payments at defaults) and Done only on the survey stage and on calibration
  3. Scoping the project: data sources, the gap in four layers, three phases, example questions, questions for NEF

Sources

Every figure is pinned in content.ts: a URL and publication date for HMRC, Budget 2026, OBR, ONS and UK Data Service sources; a repo, commit and path for policyengine-uk (98381bf), microcosm (bcb5cba) and policyengine-uk-data (ed19e33); and the calibration dashboard for the certified build of 15 November 2026. SPEAKER-NOTES.md says what each status rests on and lists what could not be sourced; README.md lists the sources.

Shared files

lib/types.ts and components/core/SlideshowViewer.tsx carry the counter: "slides" and hideControlsHint options exactly as on vahid/nef-2026-obr-macro (#118). lib/types.test.ts has the same additions minus one test that needs the nef-2026 deck registered, so whichever of the two PRs merges second has a small conflict in that file to resolve by keeping #118's version.

Review pass, 18 November (ee4d40a, 49e29c8)

Every slide was captured at 1920x1080, 1440x810 and 1280x720 from a production build and looked at; the smallest gap between content and the footer is 54px at 1080-equivalent height (slide 8 at 1920x1080).

The four fact-check leftovers:

  • Company cars read "Partly done" in this description (above), as on slides 8 and 9
  • The 1,535-target note: slide 8 shows 1,534, the count in the pinned build of 15 November; the notes say the build of 18 November has 1,535. Both counts were re-read from the dashboard's data route in the overnight pass below
  • Static-score caveat on the options slide (13): the output row is labelled "Output: static scores", and phases 1 and 3 say the score is expected below the Budget's score and the official costing. "Static" is defined on slide 12, where it first appears. The wording is "expected", because nothing has been run. The notes give the Microcosm note's own reason for cars (the size of the list-price base) and leave the pension cap's open
  • Stage number: company_cars is the 19th of the 56 entries in the stages list of uk/source_stages.json at microcosm bcb5cba, counted from 1 (index 18 from 0). content.ts said 18th; it and the notes now say 19th. The slide shows only "56"

Also changed:

  • Slide 6: the cap chip no longer breaks between "+" and "£2,920m"
  • Slide 12: the third validation item reads as a check ("Static pension cap score expected below the Treasury's costing")
  • Slide 13: phase 3's effort note is one line ("Only pensions done"), so the three effort rows line up
  • Slide 14: "Static result is below the official costing" is now "The static score is expected below the official costing"
  • Notes: the chips on slide 6 are quoted as the slide shows them

Overnight review, 18–19 November (039933c, 48655dd)

Two full cycles. Each one captured all 16 slides' final state at 1920x1080, 1440x810 and 1280x720 from a production build and looked at every capture. A pixel scan measured the gap between content and the footer: the smallest is 52px (slides 5 and 14 at 1440x810, slide 8 at 1920x1080), so nothing touches the footer. Every number in content.ts was then checked again against its pinned source:

  • HMRC: the July 2026 Tables 1–5 (Table 5: £9,500m and £1,310m, labelled experimental; Table 2D: 51% fully electric) and the vendored Table 1 rows (920,000 cars, £3,070m). Table 4.5 for 2016-17: the shares 55/23/8/2/2/1 are of £8,250m. The tax relief statistics of January 2026: cycle to work £130m, childcare £180m in 2025-26.
  • Budget documents: Budget 2026 Table 6.1 row 86 (+£310m, +£345m) and Table 6.4 rows 14 and 15 (+£5,350m and +£2,920m; +£700m and +£975m). The Annex A rates (4% to 10%, the 37% maximum to 2027-28). Budget 2025 costings pp. 47 and 72.
  • Guidance and surveys: the OBR note of 5 February 2026 (the relief-at-source timing effect), the employer rates page (Class 1A 15%), the payrolling policy paper, the booklet 480 Appendix 12 exemptions, and FRS question ExpBen (13 benefits). The SPI field EPB and ASHE 2026 (provisional, 22 October 2026).
  • Code: policyengine-uk at 98381bf (53 YAML cases, 24+10+11+8; what is not modelled; the cap test's £1.8bn ± £1.0bn); microcosm at bcb5cba (56 stages, company_cars the 19th; the FRS items; the defaults); policyengine-uk-data at ed19e33.
  • Calibration dashboard: the dashboard's data route, for the pinned 15 November build. It has 1,534 targets, all included. The 5 company-car targets and the 7 salary-sacrifice targets are each within 1% (the worst is diesel, −0.54%). The 18 November build has 1,535.

Fixed:

  • Slide 5: the car count now reads "may be an undercount". HMRC's commentary (section 8) says the 920,000 "may be an underestimate"; the slide had said it is one
  • Sourcing: the 2026-27 car fuel multiplier (£29,200) and the van charges in the notes are sourced to SI 2025/1254 (in content.ts, the README and the notes). Budget 2026 Annex A, which content.ts had cited, doesn't list them
  • Notes: the FRS also asks the value of private fuel and of other vouchers. The notes had said it gives amounts only for pension salary sacrifice and childcare vouchers
  • Slide 12: "the official costing" for the pension cap, as on slides 13 and 14 (it said "the Treasury's costing" here only). The SPI item reads "benefits and expenses in one amount", matching slide 11

Decision log (Opus 5.5 (Fable limit)):

  • Screenshots came from a production build on a spare port (3012). In headless Chrome the dev server on 3002 showed Next's dev error overlay over the slide, so it couldn't be used for captures. The 3002 dev server was left running
  • The section slides' title stays fixed-size (the shared header pattern, as on Add NEF deck on the OBR macroeconomic model #118), so at 1920x1080 they read sparser than at 1440x810. Left as is: it is a template-wide choice, not a defect of this deck
  • Slide 5 repeats the definitions from slide 4 in its "In plain words" column. Kept, because the size slide is where Class 1A and taxable value carry numbers
  • To check them, these published files were read into a scratch folder: HMRC's Tables 1–5 and Table 4.5, Budget 2026 Table 6.4, the Budget 2025 costings, the tax relief tables, the FRS and SPI documentation, and the OBR note

Merge order: #119 targets main and is not stacked on #118. Either can merge first; whichever merges second resolves the small lib/types.test.ts conflict by keeping #118's version (see Shared files).

For Vahid:

  • The PR discussion holds no list of leftovers (only this description and the Vercel bot), so the four items above follow the task's wording
  • At 1280x720 the body text is small against the header, as on Add NEF deck on the OBR macroeconomic model #118's deck: the header keeps its size while the body follows the viewport
  • Commit trailer: the task asked for Co-Authored-By: Claude Opus 5.5; the session's attribution rule gives Claude Fable 5.2, the model that wrote these commits, so they carry that

Checks

  • On 48655dd: bun --bun run typecheck, lint (0 errors; 34 warnings, none in this deck, the same as before), test (457 passed), build
  • CI on 48655dd: Build, Lint, Tests and the Vercel preview pass
  • PDF: ~/nef-slides/nef-bik-2026.pdf, exported from a local production build of 48655dd with bun scripts/export-pdf.mjs ... --url (node isn't on the PATH). 16 pages
  • The repo's CLAUDE.md asks for GitNexus impact checks; those tools weren't available. Every change in both passes is under slideshows/nef-bik-2026/ (content, notes, README and one string in slides/index.tsx)

Round 2: cross-consistency (Opus 5.5 (Fable limit), round 2)

  • Spot-check: 15 figures checked against their pinned sources, all OK. They are HMRC Tables 1–5 and 4.5, Budget 2026 Tables 6.1 and 6.4, policyengine-uk 98381bf, microcosm bcb5cba, the calibration dashboard for the 15 Nov build, and the FRS ExpBen codes.
  • Not checked against the SI: the fuel and van figures were matched to HMRC's information note and the parameter file, not to SI 2025/1254's text.
  • No change.

🤖 Generated with Claude Code

A 16-slide deck at /slides/nef-bik-2026 for a scoping meeting with the New
Economics Foundation on modelling employer-provided benefits in kind in
PolicyEngine UK: what a benefit in kind is and how big, what the model
covers today (company cars, car fuel, Class 1A on both, pension salary
sacrifice), the data available, the gap, three phases and example questions.

- Every figure is pinned in content.ts to the source it was read from: a URL
  and date for HMRC, Budget, OBR, ONS and UK Data Service pages; a repo,
  commit and path for policyengine-uk, microcosm and policyengine-uk-data;
  the calibration dashboard for the certified build of 15 November 2026.
- Status chips (Done, Partly done, Not yet) on each pipeline stage and each
  row of the coverage map, with a key; SPEAKER-NOTES.md says what each
  rests on.
- Viewer: brings over counter: "slides" and hideControlsHint from the NEF
  macro deck's branch (lib/types.ts, components/core/SlideshowViewer.tsx,
  lib/types.test.ts), without the one test that needs that deck registered.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@vercel

vercel Bot commented Nov 18, 2026 •

Copy link
Copy Markdown

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

Project Deployment Actions Updated
policyengine-slides Ready Ready Preview Nov 21, 2026 11:20pm UTC

Request Review

- Title: "Modelling benefits in kind in PolicyEngine UK: scoping with NEF".
- Size slide: the £9.5bn is the taxable value of benefits liable to Class 1A
  (experimental); the 920,000 is what is reported to HMRC, an undercount.
- Timeline: the 10% rate for 2030-31, mandatory payrolling and the car
  ownership scheme change are marked "Announced"; the yield chips say which
  rates they cost; the pension cap is relief on the first £2,000 a year.
- Coverage map: company car rules and data are "Partly done", as on the
  pipeline slide; the calibration fact names the 15 November build.
- Data table: ASHE is "Possible; access unconfirmed"; HMRC's 2016-17 split
  by type is for validation.
- Examples and phases: Class 1A results are revenue only; the static result
  for the pension cap is below the official costing; effort is a rough
  ranking.
- Notes: what was announced and what is legislated, the stale text in
  microcosm, the SPI licence condition on publishing.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@vahid-ahmadi vahid-ahmadi changed the title Add PolicyEngine–NEF benefits-in-kind collaboration deck Add benefits-in-kind scoping deck for NEF Nov 18, 2026
- Options slide: outputs are labelled static scores, and the two that can
  be checked are worded as expected below the official figures
- Gap slide: "static" defined where it first appears; the pension cap
  check reads as a check
- Example questions: the cap's static score is "expected" below the
  costing, since nothing has been run
- company_cars is the 19th of 56 stages in source_stages.json, not the 18th
- Timeline chip keeps the sign with its figure; notes quote the chips as
  the slide shows them

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
The notes gave a cause for static scores sitting below official costings
that no source states. For cars they now give the Microcosm note's own
reason (the size of the list-price base); for the pension cap the cause
is left open.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
- Slide 5: the company-car count "may be an undercount", HMRC's own
  wording ("may be an underestimate", commentary section 8), not "an
  undercount"
- content.ts, README: the 2026-27 car fuel multiplier is sourced to the
  Van Benefit and Car and Van Fuel Benefit Order 2025 (SI 2025/1254),
  not Budget 2026 Annex A, which doesn't list it
- Notes: van and fuel charges sourced to the same Order; the FRS also
  asks the value of private fuel and other vouchers; the calibration
  counts (1,534 targets in the 15 November build, 1,535 on 18 November)
  re-read from the dashboard's data route

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Slide 12: "the official costing" for the pension cap, as on slides 13
  and 14 (it was "the Treasury's costing" here only)
- Slide 12: "SPI: benefits and expenses in one amount", matching what
  slide 11 says the SPI field holds

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

@MaxGhenis MaxGhenis left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks Vahid. The figures on slides 5, 6 and 8 match HMRC's July 2026 tables, Budget 2026 and the 15 November build. I'd make the first seven changes before 11:15, since they change what NEF sees; the rest can wait if time runs short.

Slides, before the call

  1. Slide 11: Microcosm UK builds on the FRS (frs_spine, stage 1 of 56), so its Use chip should read "Base dataset" rather than "Impute". The SPI row keeps "Impute".

  2. Slide 12: add a fifth layer, Behaviour. In the Budget 2025 costing of the £2,000 salary sacrifice cap, HMRC and the OBR take the 2030-31 yield from £5,070m static to £2,585m after behaviour (policy costings, pp. 47-48). The OBR's note of 5 February 2026 (Table 1.6) splits the 2030-31 drop into employers switching to ordinary contributions (£0.7bn), pass-through to wages and profits (£0.7bn), employees moving to relief-at-source schemes (£0.6bn, partly timing) and other responses (£0.3bn). The 2030-31 car costing also nets off drivers who switch to cheaper cars. Something like:

    • Pension cap: £5,070m static, £2,585m after responses in 2030-31 (Budget 2025)
    • Employers passing costs on to pay: modelled for the pension cap only
    • Car choice after rate rises

    The slide's grid holds four cards, so this takes five narrower columns or a three-and-two layout, and the title and notes heading become "The gap, in five layers". Worth a fresh export of slide 12 afterwards.

  3. Slides 12-15: drop the "expected below" claims and the "static" label on the pension cap. Our cap run doesn't hold behaviour fixed: whenever the cap applies, policyengine-uk cuts every worker's employment income by 0.16% (gov.contrib.behavioral_responses.salary_sacrifice_broad_base_haircut_rate). The test behind the £1.8bn doesn't use Microcosm UK, and the £5,350m and £2,920m we set against it already include behaviour, where the Budget 2025 static costing gives £5,070m for 2030-31. I'd word them:

    • Slide 12: "Compare a static car score with the official costing" and "Compare the pension cap score with the official costing"
    • Slide 13: "Output" as the row label, and the phase 1 and phase 3 outputs without their "Expected below…" sentences
    • Slide 14, second question: "NICs relief only on the first £2,000 a year, against no cap"
    • Slide 15: "Which responses matter most to you: pay restructuring, scheme switching, car choice?" in place of "Static results, or with behaviour?"
  4. Slide 6: Finance Act 2026 s. 17 already brings employee car and van ownership schemes into the benefit rules from 2030-31, so that event loses its Announced chip. The 10% rate keeps it.

  5. Slides 9 and 12: the FRS asks about both loans and accommodation: whether the home goes with someone's job (AccJob), whether the landlord employs someone in the household, and loans from an employer (OthDed, DLOANTY). It lacks the values. Slide 9: "FRS asks who, not the value"; slide 12 Data: "No values for medical insurance, loans or accommodation in the FRS". The slide 9 notes need the same change.

  6. Slide 5: HMRC charged the 2024-25 liability at 13.8%, so 15% of £9.5bn doesn't give the £1.31bn beside it. Suggest "13.8% in 2024-25, 15% since April 2025". Also "fully electric", and put the Experimental label on both Table 5 figures.

  7. Slides 11 and 12: HMRC's Table 4.5 gives provisional 2017-18 figures by type. Slide 12 Calibration: "No final HMRC totals for other benefit types since 2016-17"; slide 11 Latest: "2016-17, last final".

Slides, if time allows

  1. Slide 9, car fuel: the code doesn't model part-year fuel, and the fuel benefit reads the same imputed CO2 and default availability that leave cars at Partly done. Suggest Partly done for rules and data, with "No part-year fuel; CO2 and availability as for cars".
  2. Slide 11, ASHE: employers report the total value of all benefits in kind for each job, the only record-level value on the slide. Suggest "Value of all benefits in kind per job; pension salary sacrifice", with the title "Data: surveys mostly say who, HMRC says how much".
  3. Slide 13, phase 2 output: "Reforms to medical insurance, loans or accommodation, or to Class 1A (revenue only)". Phase 2 doesn't reach every benefit.
  4. Slide 4: only the first OpRA chip says "Exempt", so a reader can take the other three as things OpRA taxes. Suggest an "Exempt from OpRA" label and four plain chips.
  5. Slide 16: drop the github.com/PolicyEngine/microcosm link; its README still uses the old name.

Getting it on screen

  • /slides/nef-bik-2026 reaches policyengine.org only once this merges, and Vercel asks for a login on the policy-engine team preview. The max-ghenis-projects preview in the Vercel comment opens publicly, so present from it if the merge doesn't land in time.
  • #118 needs to go live for 12:00 too. Whichever PR merges second conflicts in lib/types.test.ts: keep #118's file and widen its NEF-deck test to both decks (it.each(["nef-2026", "nef-bik-2026"])).
  • In a 1280x720 share, body text on slides 8, 9 and 13 gets small, so share a full-screen 1920x1080 window.

Speaker notes

  • We've scored the 2030-31 car rise on the 3 November national, and both runs come in below the Budget's figures (PolicyEngine/policyengine.py#731, PolicyEngine/uk-autumn-budget-dashboard-2026#392). Drop "nothing has been run" and "30% to 45% short", and say "we've scored it, it comes in below, and we haven't established why". Keep the figures out of the slides and the notes.
  • The £1.8bn comes from a regression pin set in February (±£1.0bn), so the test passes anywhere from £0.8bn to £2.8bn.
  • The slide 8 line about the older UK data file can go, since the slide covers Microcosm UK.
  • hmrc_benefits_in_kind_facts.json holds current text: its rows seed the stage's draws, and the calibration takes the same figures from Chronicle.
  • "Not published after 2016-17" holds for final figures only, since Table 4.5 gives provisional 2017-18 figures.
  • If NEF asks how £9.5bn compares with 2016-17: HMRC's Table 5 gives £9.1bn that year, against the P11D-only £8.25bn.
  • Class 1A on cars and fuel came to £440m of the £1.31bn, so the model's Class 1A covers about a third.
  • The SPI licence covers seminar presentations and other distribution as well as publication (UK Data Service SN 9638 conditions of use).
  • The notes for slides 6 and 11 still credit the cap costing to the Treasury, where slide 12 now says "official costing". The OBR's note sources it to HMRC and the OBR.
Smaller notes and after-the-call code
  • Slides 11 and 13: the Use and Effort chips reuse the Done, Partly done and Not yet colours from slides 8-9, so "Validate" can read as Not yet. One neutral chip style would avoid it.
  • Slide 6: "2028–30 rates: +£700m" reads as a two-year total; "Rates to 2029-30: +£700m in 2028-29".
  • Notes for slide 6: the 40% maximum for 2030-31 comes with the announced 10%; payrolling sits in OOTLAR paragraph 1.24 alone; Parliament enacted the pension cap in the National Insurance Contributions (Employer Pensions Contributions) Act 2026, s. 1.
  • content.ts: the stage sets electric range at 50 miles for 1-50 g/km cars and leaves the 2020 registration flag at its default. Full CI passed on 329ae39, which 98381bf only re-versions.
  • Sources for slide 9's pension row: content.ts and the README cite an older imputation file, but the 15 November build takes the data from Microcosm's own salary_sacrifice stage (uk_runtime/salary_sacrifice.py).
  • The 15 November build ran microcosm 3a7067f on policyengine-uk 2.177.7. Pinning microcosmCommit to 3a7067f shows the code that ran; salary_sacrifice.py has changed since.
  • OBR salary sacrifice link: the obr.uk/download/ URL lands on "No Access" for scripted clients; the direct PDF works.
  • README, config.ts and content.ts still say the date isn't fixed.
  • PR body: drop the agent notes ("For Vahid", the commit-trailer line) before merge.
  • After the call: about 40% of deck.module.css styles #118's figures only, and BuildScope.tsx duplicates #118's. Moving the shared parts into components/ and documenting the data-step pattern in CLAUDE.md would stop the next deck copying both.
  • Slide 13: Flow renders its chevrons without data-step, so the first build may show arrows pointing at empty cards; worth a look.

Slides:
- Slide 11: FRS Use chip "Base dataset"; ASHE measures the value of all
  benefits in kind per job; HMRC by type "2016-17, last final"; title
  "surveys mostly say who"; Use chips in one neutral style.
- Slide 12: a fifth layer, Behaviour (Budget 2025 cap costing, £5,070m
  static and £2,585m after responses in 2030-31); title "five layers";
  FRS asks about loans and accommodation but lacks their value; "No final
  HMRC totals for other benefit types since 2016-17"; Validation compares
  scores with the official costings instead of expecting them below.
- Slides 13-15: drop "expected below" and the "static" label on the
  pension cap (policyengine-uk cuts all pay 0.16% when the cap applies);
  phase 2 output names its benefits; behaviour question reworded.
- Slide 6: car and van ownership schemes are law (Finance Act 2026
  s. 17), so no Announced chip; "Rates to 2029-30: +£700m".
- Slide 5: Class 1A 13.8% in 2024-25, 15% since April 2025; Experimental
  on both Table 5 figures; "fully electric".
- Slide 9: car fuel Partly done (no part-year fuel; CO2 and availability
  as for cars); loans and accommodation "FRS asks who, not the value".
- Slide 4: "Exempt from OpRA" label over four plain chips.
- Slide 13: each arrow appears with the card after it.
- Slide 16: drop the microcosm link.
- Meeting date 19 November 2026 on the cover and footer.

Speaker notes and sources: drop "nothing has been run" and the 30-45%
expectation (both car runs come in below the Budget; cause not yet
established); the £1.8bn is a February regression pin on an older data
file; behaviour figures from the OBR's Table 1.6; the facts file isn't
stale; legislation for the pension cap and car ownership schemes; SPI
licence covers seminars; cite Microcosm's salary_sacrifice stage.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Slides:
- Slide 12 Behaviour card: "Employers passing costs on to pay: we model
  it for the pension cap, not for benefits in kind" (policyengine-uk
  passes employer Class 1 NICs on to pay by default, but not Class 1A);
  keep "2030-31" on one line.
- Slide 6: "Rates to 2029-30: +£700m in 2028-29".
- Slide 4: "Exempt from OpRA" label keeps the acronym's case.
- Slide 16: contacts in one row, so the policyengine-uk label doesn't
  wrap.

Notes and sources:
- Slide 13: the pension cap gap is open on Microcosm UK too (an October
  run lacked bonus sacrifice, microcosm#1340; its base still falls short,
  microcosm#1390).
- Slide 12: why slide 6's £2,920m differs from Budget 2025's £2,585m;
  how policyengine-uk handles pay pass-through; the £1.8bn pin is for
  2030-31.
- Slide 6: the cap's NICs charge needs regulations under SSCBA s. 4(6A),
  none made by 18 November 2026.
- Slide 14: which examples have an official costing to check against.
- Slides 5, 11, 15: wording and scope fixes (ASHE's record-level total,
  pass-through to wages and profits, who acts in each sentence).
- Drop the calibration dashboard link and use the OBR note's direct PDF;
  correct the content.ts comments on electric range and CI.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… on to pay

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@MaxGhenis

MaxGhenis commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

I've pushed the fixes from my review (0cc4e29, 58d8351, 4edbd4f, 48c0298), so please pull before pushing anything else. Everything's in except pinning microcosmCommit to 3a7067f, which can wait until after the call. Present from https://policyengine-slides-ngyphtybf-max-ghenis-projects.vercel.app/slides/nef-bik-2026 unless this merges first.

- Notes: the pension cap's only pay effect is the 0.16% cut; correct the
  employer NICs pass-through wording (slides 12 and 13)
- Notes: relief-at-source £0.6bn is a temporary timing effect (OBR para 1.15)
- Notes: SPI incomes go onto copied households; seven HMRC targets, not OBR;
  drop a link to a page that shows an old name; OpRA exemptions 'include';
  globally mobile employees outside payrolling; active voice
- Slide 12: Behaviour card gives the Budget 2026 recosting beside Budget 2025
- Slides 11-13: keep 'part-year' and 'salary-sacrifice' on one line; validate
  cars by income and region
- content.ts/README: add the HMRC 2019 commentary source; unlink
  source_stages.json; accurate source-pinning description

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Main gained the BIST deck (#115), which added the same two viewer options
under different names. Take main's countSlides and hideNavHint, drop this
branch's counter and hideControlsHint (and the counterLabel helper and its
test), and switch the benefits-in-kind deck config and README to main's
names.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

This branch was successfully deployed

1 active deployment
Preview – policyengine-slides — 03568349 Deployed Nov 21, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants