Model register

1.9.1 balance fixes

• Voters partly look past the crisis when you are prepared. The size of the shock is what the crisis does to an unchanged government's standing with voters (peaceful minus crisis scenario, same half-year, every channel). You are forgiven a share of it: none below readiness 94, rising to 70% at inherited readiness 100 and up to 90% at 110 or more. It does not depend on your other policies.
• Readiness is a safeguard in the country verdict. Falling behind the unchanged path counts as harm; being above it cannot on its own make the country better off.
• Warning and time to respond. In the crisis scenario, year 3 carries an intelligence warning and the year-3 and year-4 chapter openers foreshadow the conflict. The year-4 Budget can buy urgent supplies (£4bn, +2 readiness before the crisis). Defence blame is judged from the crisis's second half-year, after the first in-tray.
• Allied assistance now costs £3bn and lasts a year, so it covers the half-year that no in-tray reaches.

Effect with the services electorate: changing nothing ends on 47.5% support in the crisis against 49.9% in peace (it was 42.9%). Of 450 random strategies, 17% are re-elected in the crisis against 40% in peace, and all four endings are reachable. For a −20% defence cut, answering the crisis with Budget procurement and allied help cuts the competence penalty from 7.3 to 4.2 points.

Current model: version 1.6 realism update

Version 1.6.0, 26 September 2026. This release fixes the problems found in the model review. It is the current specification wherever it differs from the first-edition register or the growth notes. All coefficients remain provisional game assumptions unless a source is named. They are chosen to give the right direction and rough size, not fitted estimates.

What changed, in plain words

1. Reckless borrowing now ends badly. Markets, inflation, the Bank of England and voters all push back, and a big enough loosening sets off a gilt crisis.
2. The market reaction has a more realistic shape. Small giveaways cost a little. A surprise the size of the 2022 mini-budget causes a jump that fades. Very large, lasting gaps become steeply more expensive. Debt levels count, and borrowing to invest worries markets less.
3. Inflation is part of the model. It rises with excess demand, VAT rises and a falling pound. The Bank of England reacts. Prices feed through to receipts, benefits, budgets, debt interest and household incomes.
4. Index-linked debt costs the right amount. It now follows the model's own inflation rate. The old version was about 7× too small.
5. Rate changes reach the interest bill faster. A quarter of debt pays Bank Rate straight away, and new issuance pays market rates.
6. Homes and transport are lasting assets. They no longer disappear when funding returns to normal.
7. Taxes behave more realistically. Big rises raise less per point, the additional rate has a revenue peak, cuts cost the full amount, and receipts arrive with a collection lag.
8. The devolved share is included. Scotland, Wales and Northern Ireland funding is now added to all English service spending, not just the NHS.
9. Slow things are slow. Skills take decades. NHS gains need staff and show diminishing returns, while cuts bite faster. Police and courts take about 18 months to respond.
10. New supply-side effects. Lower income tax and NI slightly encourage work, better NHS access gets people back to work, and corporation tax partly falls on shareholders.
11. Voter drift is fixed. Voters no longer reward the inherited income trend or punish the inherited rise in interest costs.
12. Two new levers: VAT and employee National Insurance.
13. The doctors' payroll is corrected from £10bn to £6.5bn.

Before and after

Deterministic runs with the services electorate. Policies are set in year 1 and held. External events are off unless stated.

Game balance note. With the authored events switched on, a Chancellor who changes nothing now finishes just below the line (49.8%) instead of just above it (53.3%). This happens because voters no longer get free credit for the inherited income trend, and they now dislike the energy-shock inflation. If you want "doing nothing" to win narrowly, raise the starting support. That is a design choice rather than a realism fix.

Markets

• Structural gap G = ongoing day-to-day policy cost (NHS, benefits, justice, education, separately funded pay) − full tax yield + interest on extra debt (at the rates locked in when it was borrowed or repaid), plus half of investment (housing, transport, the diagnostic programme).
• g = G as % of nominal GDP. d = extra debt versus the inherited path, as % of GDP.
• Target premium = 0.25g + 0.15 × max(0, g − 1.5)² + 0.03d. It has a floor of −0.5 points and a ceiling of 12 points (loss of market access).
• Rises in the premium apply at once. Earned reductions: half arrives at once, then 25% of the remaining gap each half-year.
• Surprise jump: a single announcement that loosens G by more than 0.75% of GDP adds 0.8 × the excess as a temporary jump. The jump halves every six months. Only loosening that pushes G above the inherited path counts, so moving from a large surplus back towards balance causes no jump. (1.6.1)
• Gilt crisis: premium + jump ≥ 2 points triggers it, once per term. Effects:
• credibility −8 with all voter groups for the rest of the term
• a confidence shock of −1.5% of output, fading by 40% each half-year
• sterling falls by 1.5% + 0.6% per point of stress above 1
• Financial stress drag: output falls by 0.25 × (stress − 1) while stress is above 1 point.
• Borrowing rate = inherited gilt rate + external shock + premium + jump + 0.5 × Bank Rate deviation.
• Each half-year, debt is fixed-rate once issued. Extra borrowing, or repayment through surpluses, is locked at the inherited rate of its period. Everything actually issued (4% of inherited debt refinancing plus actual borrowing) also pays that period's market spread, from the next half-year. So a jump in the rate on new borrowing reaches the interest bill gradually, and higher rates can never lower it. Results also show the average rate paid across all debt. (1.6.1)

Scale checks: an unfunded 1% of GDP loosening costs about 0.25 points in the long run. Evidence suggests roughly 0.1–0.3 points. A surprise of about 1.5% of GDP gives a jump of about 1 point that mostly fades within a year. In 2022, gilts moved about 1 point at the peak, and much of that was pension-fund selling.

Prices and the Bank of England

• Demand pressure = 0.7 × previous + 0.15 × gap + 0.05 × max(0, gap)² + 0.3 × sterling inflation. The gap is temporary demand output net of stress drag and confidence.
• Inflation = 2% + energy (9 × shock + persistent energy pressure) + demand pressure + VAT pass-through + sterling pass-through.
• VAT: each point lifts the price level 0.45%, spread over about a year.
• Sterling: the price level moves 50% towards the sterling target each half-year.
• Bank Rate deviation. It moves halfway each half-year towards 1.5 × (demand pressure + sterling inflation) + 0.5 × gap, with a floor of −3.
• It looks through the authored energy shock, because the scheduled 1-point external yield rise stands for that tightening.
• It also looks through one-off VAT changes.
• Price levels. CPI accumulates the extra inflation. The domestic GDP deflator takes demand pressure plus half of imported inflation (energy, sterling).
• Budget effects, using the price level at the start of each half-year:
• receipts rise 1.1 × the domestic price level, including fiscal drag from frozen thresholds
• benefits are uprated with last year's CPI
• departmental budgets receive half of last year's extra inflation; the rest erodes real service funding
• Debt interest:
• a quarter of debt (QE reserves and Treasury bills) pays Bank Rate at once
• 24% is index-linked and pays the extra inflation each period
• Households:
• demand-driven inflation squeezes real incomes point for point until wages catch up
• a weaker pound costs 0.8 × its price effect
• mortgage-rate changes (0.7 × Bank Rate + 0.5 × gilt stress) hit the lower, middle and higher income groups by −0.3, −0.8 and −0.2 per point
• Debt/GDP uses nominal GDP, so inflation lowers the ratio, as it does in reality.

Taxes

HMRC illustrative tax changes (January 2025 bulletin), £bn per point:

• Yield of a rise of x points: mature × (x − x²/2T). Cuts are linear, because the HMRC costings already include behaviour.
• Scaling and timing. Yields scale with nominal GDP from the 2027–28 anchor. Receipts use the collected rate, which moves the stated share of the way to the set rate each half-year. This reproduces HMRC's first-year versus mature ratio.
• What the market and voters see. The market gap uses the full mature yield. Households feel income-tax, NI and VAT changes at once. Corporation tax falls on them gradually, mostly on higher incomes.
• Income effects per point (lower / middle / higher):

• Labour supply, as % of output per point: basic 0.04, NICs 0.04, higher 0.02, VAT 0.01. It adjusts 25% each half-year. Better NHS access adds 0.01% per access point.
• Tax-focused voters now also weigh NICs (0.45 × points/3) and VAT (0.35 × points/2).
• Preferences saturate: 12 × tanh(raw/12).

Services

• Devolved share added to the cost of English spending (full Barnett comparability ≈ 1.19): education 1.19, housing 1.19, transport 1.15, police/courts/prisons 1.13. Chart slices still show the English department. The devolved share sits in "Other".
• Housing and transport stock. Each half-year: asset = asset × (1 − 0.06) + 0.2 × (delivered funding − real erosion). Target index = 100 + 0.8 × asset. So +20% funding held for eight years gives about +34 points and stays well above 100 for years after it stops.
• Transport productivity is now 0.01% output per index point: about 0.35% of GDP for eight years of +20%. That's consistent with public-capital output elasticities of about 0.1.
• Police, courts and prisons: the lag is now three half-years.
• Skills: move 2.5% of the way to education capacity each half-year (a half-life of about 14 years).
• NHS:
• target effect of funding change x = 0.85x(1 − x/60) for increases and 0.85x(1 + |x|/40) for cuts, minus 0.85 × real erosion
• capacity rises 20% of the way to target each half-year and falls 40%
• Doctors' pay: about 81,000 resident doctors in England (Full Fact, January 2026). The illustrative employer pay bill including on-costs is £6.5bn, so a 30% settlement costs about £2bn a year. Campaigns saved before 1.6 keep their £10bn payroll.

Politics

• Material = 2 × delivery + 20 × [0.45 tanh((income − 100 − inherited trend)/5) + 0.2 tanh(employment/2) + 0.35 tanh((NHS access − 100)/10)], then two deductions:
• inflation: 6 × tanh(max(0, inflation − 2.5)/2.5)
• market stress: 4 × tanh(max(0, premium + jump − 1)/2)
• Fiscal worry = −0.5 tanh((interest/receipts − inherited ratio for the same year)/0.03) − 0.3 tanh(G/GDP/0.02) − 0.4 tanh(stress/1.5). It is weighted 2 / 4 / 10 across the groups.
• Gilt-crisis credibility: −8 for every group.

Not yet modelled

• the pensions triple lock as a lever
• personal-allowance and threshold changes as levers (frozen thresholds act only through fiscal drag)
• employer National Insurance
• a full wage–price model
• an explicit NHS workforce series
• regional detail

Saves

Version 1.5 and earlier saves migrate automatically: - new taxes start at zero, and their collected rates equal the current policy - housing and transport assets are set from current delivery - justice queues are padded with the current setting - price and market fields start at zero - the original £10bn payroll is kept

Closed accounts are never rewritten.

World events (2.0.0)

Each term draws its own world from a seed. Same seed, same world: the “changed nothing” comparison, “what they cheered vs what worked”, replays and shared links all use the saved schedule. Events are fictional scenarios using the model's existing channels, not predictions.

• Year 1: one opening pay dispute: resident doctors (30% claim, £6.5bn payroll, NHS backlog), teachers (10%, £22bn, education capacity) or rail workers (15%, £4bn, transport capacity and a little output). Absorbed pay squeezes that service like a funding cut.
• Years 2–5: three events, one per year, at 80–120% of base size and in either half-year. At most one price shock (energy, trade war, security crisis) and one market shock (borrowing costs, financial crash). A commodity windfall appears in about a third of worlds. The security setting can force the crisis in, keep it out, or let the deck decide.
• Fairness: a difficulty budget limits the bad news, and a world is kept only if a government that changes nothing ends between 47% and 51.5% support.

The first choice listed is what an unchanged government does. Saves from before 2.0 keep the original fixed world (doctors, energy in year 2, borrowing costs in year 3 and, unless peaceful, the security crisis in year 4).

The two verdicts (1.9.0)

Military readiness is now a separate measure with a three-point tolerance; materially weaker readiness counts as harm alongside the measures below.

The voters decide on election night: more than 50% support wins. The country is judged ten years on, with your final policies carried forward, against a term in which you changed nothing and faced the same external shocks. The game's view is that the public judges what it feels now, while the long-term result is what matters for the country.

Better off means either (a) at least one of the first four measures is better, none is worse, and debt is not worse; or (b) debt is better and none of the first four is worse. So improving lives by borrowing does not count, and neither does repairing the debt by making lives worse. Anything smaller than the tolerances counts as about the same, so changing nothing is never better off.

What they cheered vs what worked replays your term from its recorded decisions with one lever held at its inherited setting, for each of your biggest lasting changes (up to four, ranked by their largest annual effect on the gap). “At the time” is the change in support at the end of the year you first made it; “ten years on” applies the same measures. Effects interact, so rows do not add up.

Balance check (900 random five-year strategies): tax cuts with spending cuts reach “better off” in 9% of cases, tax rises with spending rises in 18%, unfunded tax cuts in 1%. Every broad approach can reach every ending; the model's growth effects from tax cuts are modest, which is a stated limitation.

Consequences (1.7.0)

Consequences are deterministic game thresholds, not estimates of real-world crisis probabilities. The same rules run in draft forecasts, at the mid-year in-tray and in year-end results.

Tiers

The gilt crisis (stress ≥ 2, once per term) and a broken fiscal rule are always severe. The energy shock is shown as “outside your control” with its size and its effect on lower incomes (about 15 points per unit of shock) and inflation (about 9 points). Good news: a surplus, debt falling as a share of GDP, a market discount, a fiscal rule holding, NHS access above 104, a service above 104, and any pressure that was at warning or worse last year and has now cleared.

Causes, streaks and blame

Each event lists its numbers and its main causes with their size, and names the levers behind it (used for the warnings beside each control). A snapshot of causes is stored with each closed year. Titles change in the second and third consecutive year at warning or worse. A severe pressure in its third year running costs 0.75 support points with every group for the following year (at most −3 in total); the cost is recalculated each year, so fixing the problem removes it.

Strikes have mechanical effects: at warning each half-year adds 0.5 to the NHS backlog and takes 0.5 off education and justice capacity; severe doubles this. Councils have no separate budget in the model and are a proxy for real-terms erosion and housing cuts.

Responses

A pressure at warning or worse, either now or at last year's close, unlocks its responses at the next in-tray. One response per in-tray.

• One-off spending: household support £4bn (+2 lower-income points); NHS capacity £3bn; housing £1.5bn (+3); transport £2bn (+3); justice £1.5bn (+4); education £2bn (+2); councils £1.5bn (+2 housing); public-sector pay award £3bn (no strike disruption that half-year); energy bill discount £5bn (−1 point of inflation for six months, rebounding when it ends; +1.2/+0.8/+0.3 income points); brought-forward investment £5bn (investment demand, +1 transport).
• Lasting policy changes: basic rate +2; VAT −2; spending restraint (−3% on NHS, housing, transport, justice and education); U-turn (all rates and budgets back to last year's settings, jump halved, −1.5 support with every group per U-turn up to −3; only offered if it closes at least 0.25% of GDP of the lasting gap).
• Announcements: fiscal rule (−0.25 points on the target premium while it holds; a later loosening above 0.3% of GDP breaks it: +0.5 jump and −2 support with every group for the rest of the term); mortgage charter (+0.2/+0.5/+0.1 income points for six months).

Earlier edition register

The sections below record earlier releases. Where they differ from version 1.6 above — including the tax taper, service lags, output coefficients (transport 0.01, skills speed 2.5%), the market premium, the £10bn payroll and the former ±8% output bound — version 1.6 applies.

Version 1.4.0, 25 September 2026. This is the specification of the shipped educational MVP. The earlier implementation plan remains a design target; this register records the implemented simplifications. It is not a claim of completed econometric calibration. Fiscal accounting and source anchors are tested; behavioural coefficients and political weights are provisional scenario assumptions, exposed to players. External economic review and selectable economic parameter sets remain future work.

Sources and boundaries

Amounts are nominal £bn unless stated otherwise. The fixed vintage is the OBR March 2026 Economic and fiscal outlook, annex tables A.3, A.7 and A.9, fiscal years 2026–27 through 2030–31. Retrieved 25 September 2026. UK forecast, not outturn. Opening public sector net debt is £2,922bn. Published totals are rounded; spending is reconstructed as receipts plus borrowing rather than pretending rounded aggregates reconcile exactly. GDP used in the debt ratio is the OBR centred denominator, distinct from annual GDP used for policy scaling.

Interest uses central government debt interest net of the Asset Purchase Facility. It is an approximation to the relevant interest component within the wider public-sector totals, not a security-level reconciliation of PSND and gilt liabilities. Each baseline stock-flow adjustment is closing debt minus opening debt minus borrowing; it is carried separately, not hidden in spending.

Spending Review 2025 supplies department anchors for DHSC, NHS England, DfE and MOD. These are departmental settlements, not complete UK functional health/education/defence spending. Devolved services are in the residual. Later allocations after the published resource settlements are model extensions. Welfare and pensions use OBR welfare totals. “Other” reconciles all remaining spending, including capital and devolved additions. Chart categories must not be read as internationally comparable functional spending shares. Transport includes DfT and HS2. Policing, justice & borders combines Home Office, Ministry of Justice and Law Officers’ departmental budgets, including border and asylum spending. SR2025 table 5.1 supplies 2026–27 to 2028–29 allocations (£32.2bn/£32.0bn/£31.5bn for transport and £38.6bn/£38.6bn/£39.2bn for policing, justice & borders); the following two years assume 3.5% annual nominal growth. These slices are deducted from Other without changing total spending.

HMRC illustrative tax changes, June 2025 update supplies local annual rate-change anchors: basic income tax £6.9bn per percentage point; higher £1.6bn; additional £0.145bn; corporation tax £3.6bn for a point on both main and small-profits rates. Higher/additional rates move together. Corporation tax excludes the ring-fence base. Income-tax figures already allow for some taxable-income response. The local costings already include behaviour; the large-change taper below is an additional scenario assumption. Timing, fiscal devolution, marginal relief and interactions are simplified; annual effects are scaled by baseline nominal GDP. Beyond ±3 points of basic tax or ±5 points of higher/corporation tax, additional points produce 65% of the local marginal yield (symmetrically for cuts). This taper is an illustrative game coefficient, not an HMRC estimate for large reforms; abolition is outside the available range.

Government material is generally available under the Open Government Licence where indicated by the originating publication. The site links to originals, uses numerical aggregates and original prose, and does not use official logos or suggest government endorsement.

Accounts and timing

The engine uses two half-year periods. Each annual baseline flow is split equally (a simplification of cash collection). Spending minus receipts equals borrowing. Opening debt plus borrowing plus the explicit stock-flow adjustment equals closing debt. Policies are levels relative to the inherited path; repeating the same level does not charge a fresh increment. A budget occurs at the beginning of a year, followed by a mid-year event. Draft previews run on a copy and include scheduled shocks but no new optional response to them.

Permanent controls: basic income tax 10–35%; higher income tax 30–55% with additional tax 5 points above; corporation main rate 15–40% with small-profits rate 6 points below. NHS, working-age generosity, housing, transport, justice and education allow ±20% of baseline funding. Players can move across the entire range in one budget; fine steps are 0.5 percentage points for basic tax and 1 for other controls. Settings persist until changed. NHS cost uses NHS England operating funds × percentage change × 1.19 (illustrative devolved addition). Benefits affect an illustrative 45% of welfare, not pensions. The remaining spending path is inherited. Borrowing is the residual of the budget.

Policy consequences and emergency responses (1.2.0)

Warnings use the same thresholds in draft forecasts and realised reports. Lower-income purchasing power below 97 triggers household pressure, below 94 severe hardship. Effective NHS access below 96 triggers pressure, below 90 a service crisis. A policy borrowing premium of at least 0.75 points triggers financing pressure, at least 2 severe stress. Other service delivery indices below 94 trigger pressure, below 90 severe pressure. NHS access above 104 reports improving capacity. These are deterministic scenario thresholds, not estimated real-world probabilities. External shocks retain their scheduled timing.

Each half-year, NHS raw capacity below 95 adds 0.12 × the shortfall to the backlog. Existing backlog decays by 20%; strikes add unmet demand and capacity above 100 clears it gradually. This creates persistence after severe cuts without a permanent penalty once services recover.

At the mid-year decision, active household or NHS pressure permits one temporary response. Targeted household support costs £4bn in the second half and adds 2 points to the lower-income index for that period only. Emergency NHS capacity costs £3bn, reduces backlog by up to 2 points and raises the half-year service target by 3. The ordinary 40% adjustment toward that target applies, so capacity does not jump instantly. Both costs enter spending and borrowing once; resulting debt persists and incurs interest subsequently. They do not reset ongoing funding settings. In year two the selected response waits for the windfall stance decision before closing the half-year. No response is assumed in budget previews or legacy projections.

The budget summary compares each draft to holding current settings with the same scheduled shocks and no new optional responses. Debt detail reconciles opening debt + borrowing + inherited stock-flow adjustments = closing debt. Interest is already included in spending. Version 1.0 and 1.1 saves retain recorded accounts and migrate forward; future periods use the revised rules.

Additional service controls (1.1.0)

The pie remains a departmental allocation chart. Housing & regeneration is MHCLG Housing, Communities and Local Government capital DEL from SR2025 table 5.4: £9.8bn, £9.2bn and £9.2bn in 2026–27 through 2028–29. This is a broad investment proxy including regeneration and community infrastructure, not a claim that the whole amount buys affordable homes. The next two years grow this allocation by 3.5% nominal annually. Housing, transport and justice are taken out of Other; total baseline spending does not change.

Annual direct costs are baseline allocation × policy percentage / 100. Housing, transport and education adjust their full chart allocations. Police, courts & prisons adjusts an illustrative 75% of the combined Home Office, Justice and Law Officers’ allocation. That fraction is a model choice for an operating pool, not a sourced police budget; the other 25% stays fixed. No additional devolved multiplier is applied to these four controls. Half of each annual cost is charged each half-year. Ongoing costs enter the structural financing gap and demand immediately; chart slices move by the same direct amounts. They are levels against the baseline, not additional charges stacked each year.

All indices start at 100. Each half-year queues the current funding setting; the setting leaving the queue sets the delivery target, 100 + effect × delayed setting. Cuts work symmetrically. Reversing funding does not erase past queued decisions. These are bounded illustrative service responses, not counts of homes, crime forecasts or measured returns. Construction failure, land constraints, regional differences and detailed maintenance/asset lifetimes are not separately simulated.

Workforce skills move 8% toward the previous half-year’s education capacity. Transport adds 0.04 × its index deviation from 100 to the target output percentage level; skills add 0.06 × their deviation. The existing output adjustment still applies. Housing availability adds 0.08 / 0.04 / 0.01 times its index deviation to lower-/middle-/higher-income indices, as a simplified housing-pressure effect. Direct service spending enters the demand equation with coefficient 0.7. Economic returns are level effects, not perpetually compounding growth bonuses.

Services-focused preferences gain 0.5 × setting / maximum for each new control. Material support gains 2 × the mean of tanh((index − 100) / 10) across the four delivery indices. Fiscal penalties still apply to borrowing. Forecasts, annual results, election outcomes and the conditional year-ten legacy expose the delivery indices; the existing NHS metric remains separately labelled.

Version 1.0.0 saves migrate without changing recorded fiscal accounts or existing policy settings. New policies start at zero, delivery and skills at 100, and pipelines empty. New settings and pipelines survive save/resume and continue into years 6–10. Behavioural coefficients remain provisional game assumptions.

Financing and markets

Annual structural financing gap G is NHS cost + benefit cost + separately funded pay + annual programme commitment + new service costs − direct tax receipts + 4.5% of accumulated extra debt. Temporary oil receipts do not finance permanent commitments. During year 2, a promised levy reduces expected G by half its initial £2.4bn annual yield; the promise books no actual receipts. A budget or pay/promise announcement records the change from prior expected G. It does not repeatedly add a surprise penalty.

Let r = max(0, G / baseline GDP × 100). Target policy premium in percentage points = min(4, 0.6r + 0.3r²). Increases are immediate. Lasting financing improvements immediately deliver half of the smaller of the new target reduction and the remaining gap to that target. Unchanged targets do not repeat this benefit; temporary levy promises receive no immediate reduction. Each half-year then closes 25% of the remaining gap. This is a transparent financing-risk scenario, not an estimated bond-market law. The inherited deficit is already in the baseline yield, so a positive baseline deficit alone receives no extra penalty.

Baseline yields are [4.5, 4.8, 5.0, 5.2, 5.4]%. Add external shock and policy premium. Each half-year 4% of the opening £2,922bn inherited stock refinances, with half-period interest on that new cohort. Each cohort retains its spread. Added borrowing is approximated at the current all-in rate; it is not a full fixed-coupon maturity ladder. Extra indexation = £2,922bn × 24% exposure × energy deviation × 1.2% / 2. This models incremental inflation exposure only; baseline interest is not reapplied. No separate duplicate principal indexation is added to debt.

Economic and service coefficients

All the following are provisional game parameters, not fitted UK estimates. Their purpose is a bounded model with visible causal channels, not a claim that a specific policy must succeed or fail.

• Corporation-tax memory moves 25% towards the rate deviation each half-year. Oil-investment memory increases 0.04 per active levy half-year then decays by 4%; it reduces the additional levy base, bounded at 60% of the original base.
• Version 1.5 replaces the former direct corporation-tax output coefficient with business investment and capital. Desired capital deviation (%) = −0.4 × corporation-tax rate change (percentage points) −0.5 × additional financing rate. Capital closes 8% of its target gap each half-year; investment replaces 4% of capital plus that adjustment. Productive capacity = 0.25 × capital deviation + 0.02 × delivered transport deviation + 0.03 × accumulated skills deviation. Financing sensitivity, delivery coefficients and adjustment speed are illustrative assumptions.
• At unchanged financing, an eight-point corporation-tax cut approaches 0.8% higher output and 3.2% higher investment, compared with the historical Treasury/HMRC 2013 modelling range of 0.6–0.8% GDP and 2.5–4.5% investment. This historical benchmark is not a forecast for today's tax system. Calibration source.
• Demand responds to changes in annualised fiscal flows/GDP. OBR year 0–5 profiles are tax [0.33,0.30,0.23,0.14,0.05,0], services [0.45,0.42,0.29,0.13,0.04,0], benefits [0.60,0.57,0.43,0.23,0.07,0], investment [1,0.83,0.43,0.23,0.07,0], interpolated each half-year. Unchanged policy does not restart stimulus. Transport/housing and the diagnostic build use investment; NHS, education, justice and emergency services use services. OBR methodology.
• Output combines capacity and temporary demand, less energy/oil effects. Employment deviation remains 0.35 × output; additional annual benefit cost is −£1.5bn × employment deviation. Receipts move proportionately with output in addition to the direct tax-policy costing. These aggregate feedbacks are approximate.
• The growth panel compares with unchanged inherited policies under identical external shocks. Annual growth uses annual-average output and a baseline [1.1,1.6,1.6,1.5,1.5]% profile, approximately mapping OBR calendar years 2026–30 onto fiscal turns; subsequent years use 1.5%. OBR growth profile. Existing nominal fiscal totals are retained. This release does not model policy-driven inflation. Old saves preserve closed accounts, seed capital from the previous investment response, and do not reconstruct past demand impulses.
• NHS service target = 100 + 0.85 × NHS funding percentage − absorbed pay as percent of NHS budget + usable diagnostic benefit − 8 × energy deviation. The service index moves 40% towards target each half-year, bounded 65–125. Each unresolved strike period adds 1.6 × max(0, 1 − settlement / demand) backlog units. Backlog decays 20% per period, with an additional reduction for capacity above 100. Effective access is capacity minus backlog.
• Diagnostic benefit is 3 service points after construction, or 2.25 in the stress continuation. There is no explicit clinician labour market or waiting-list microsimulation. Staffing and other funding constraints are represented by the combined service equation, not an independently calibrated workforce series.
• Household indices start at 100. Illustrative trend real income increases 0.3 points/year. Lower/middle/higher profiles use benefits weights [0.32,0,0], basic-tax weights [−0.16,−0.48,−0.17], higher-tax weights [0,−0.05,−0.40], output weights [0.55,0.60,0.70], and energy weights [−15,−9,−5]. Aggregate weights are [0.4,0.4,0.2]. Targeted relief adds 1.7 to the lowest profile; broad relief adds [1.1,1,0.5], only while paid.
• Persistent price pressure decays 4% and adds 0.08 × energy deviation per half-year. Inflation indicator = 2 + 9 × energy deviation + price pressure. This is a stylised pressure indicator, not a CPI basket or a price-index-consistent macro model. Relief changes disposable income, not this inflation indicator.

Authored event dossiers

Politics and outcomes

Three preference groups start at 52%. Electorate mixes (services/taxes/fiscal) are [60,25,15], [25,60,15], [25,15,60] percent. Changing electorate does not change the economy. Groups react to the same material conditions but value fiscal performance and policy preferences differently. The additional service-control terms are specified above. These are fictional voter archetypes, not claims about British polling.

Material contribution = 20 × [0.5 tanh((income−100)/5) + 0.2 tanh(employment/2) + 0.3 tanh((access−100)/10)]. Service preference = 10 × [0.4 NHS/10 + 0.3 benefits/10 + 0.15 completed-programme + 0.10 higher-tax/5 + 0.05 active-levy]. Tax preference = −10 × [0.65 basic-tax/3 + 0.25 higher-tax/5 + 0.10 corporation-tax/5]. Fiscal preference starts at zero.

Fiscal condition = −0.6 tanh((interest/receipts − 109.4/1304)/0.03) − 0.4 tanh(G/GDP/0.02). Multiply it by [2,4,10] for the three groups. Add trust (promise delivered +1; broken −3; late −2), ambiguity penalty −0.5, and an announcement bounce bounded ±2 which halves each period. Group support is bounded 0–100; electorate-weighted support strictly above 50 wins. Exact ties lose.

The ending reports election and national indicators independently. There is no hidden single “good country” score and no automatic equivalence between lower debt and better welfare. The reference path faces identical shocks but makes no new interventions. This is a comparator, not an optimal strategy. Economic collapse/invasion, snap elections and a playable second term are outside this MVP.

Legacy and limitations

Years 6–10 hold final policy settings. The baseline grows nominal flows and GDP 3.5% annually, accrues borrowing into debt and assumes zero further stock-flow adjustment. External premium fades to 0.75/0.5/0.25/0/0 points. Stress keeps it at +1 and reduces diagnostic benefit by 25%; it does not rewrite realised outcomes. Neither is a forecast probability interval.

Remaining work before describing this as a high-accuracy economic simulator: empirical behavioural ranges, debt maturity calibration, consistent inflation/price levels, explicit workforce/delivery constraints, distributional microsimulation, tax-base/territorial interactions, independent economic review, user-selectable world models and wider policy/events. Non-dom rules, wealth/employer taxes and a wider set of endogenous events remain future content. The current public build is labelled a simplified first edition throughout.

Engineering validation

Pure deterministic model in site/game/engine.js; sourced anchors and policy bounds in data.js; UI in app.js. Tests cover source-baseline reproduction with shocks disabled, accounting identities, forecast purity, policy bounds, financing response/recovery, external refinancing, electorate invariance, delayed investment, event expiry, promises, legacy persistence, validated save/load and seeded strategy sweeps. npm test runs offline with Node's standard test runner. Browser and live deployment checks are recorded in docs/mvp-release.md.

Distinct crisis responses and pay restoration (1.3.0)

At each mid-year event, existing consequence thresholds unlock relevant interventions. The player chooses one response or holds course. A common response definition supplies both UI options and engine eligibility checks. Responses are recorded in annual results and survive a pending year-two stance/save. Costs are one-off spending; added debt and its interest persist. Unselected pressures are not erased.

Service boosts apply after the period's ordinary delivery adjustment. They do not change ongoing funding or accelerate queued major projects. Subsequent periods converge back toward the funding-driven target, so benefits fade under continued cuts. Housing affects purchasing power, transport affects activity, justice affects service-related support, and education affects skills only from the following period. These are disclosed scenario coefficients, not empirical intervention cost-effectiveness estimates. Existing household and NHS responses remain available under their original thresholds.

The financing option is unavailable when basic tax is already at 35%. It changes the actual policy before second-half receipts and household outcomes are calculated. Expectations are repriced with half of a newly earned reduction applied immediately and the remainder gradually; a small correction may leave the premium at its cap. It is not a free confidence bonus or a promise of future tax revenue.

New campaigns use a fictional resident-doctor pay-restoration demand of 30%, with 0%, 15% and 30% settlement options. This is inspired by the scale of historical restoration claims, not a statement of current union demands or an agreed annual award. A DHSC letter of 17 April 2023 records the BMA's historical 35% restoration starting position. The June 2026 government offer describes different, staged pay awards; it is not a 30% annual settlement.

The illustrative covered payroll remains £10bn, growing 2% annually. A 30% settlement therefore costs £1.5bn in the first six months and £3bn over a full year before payroll growth. A separately funded top-up adds the existing 19% devolved multiplier (£1.785bn initially); absorption reallocates the existing NHS envelope and reduces service capacity. Pay persists. The compromise halves authored strike disruption; the full settlement ends future authored disruption without clearing past backlog. Unresolved year-two settlements use the same demand and growing payroll. Saves from 1.0–1.2 retain their original 4% demand and actual pay levels, with no retrospective rewriting of accounts.

Version 1.4 financing update

Financing improvements below the inherited path now earn a target discount: 0.3 × the financing improvement as a percentage of GDP, capped at 0.75 percentage points. As of version 1.4.1, lasting improvements deliver half of each newly earned rate reduction immediately, limited to the remaining gap to target. Each half-year closes 25% of the remaining gap towards a lower premium or discount. Unchanged targets and temporary levy promises do not earn an immediate reduction. Worsening financing still reprices immediately. External rate shocks are separate. This is an illustrative game coefficient, not an empirical estimate. Existing saves retain closed accounts; subsequent periods use the revised model.