How the model works

1. What this model is — and what it isn't

Britain 2036 is a structural scenario tool, not a forecasting engine. It models how policy changes propagate through interconnected systems — healthcare, housing, labour markets, fiscal balance, demographics — using a directed acyclic graph (with limited feedback loops) of causal relationships.

The output is not a prediction of what will happen. It is a projection of what the model's causal structure implies would happen if its assumptions hold. Those assumptions are documented in full below.

2. Four-layer architecture

The simulator operates through four distinct layers. Each layer serves a different epistemological function — separating what we know, what we model, what we're uncertain about, and what we assume.

Layer 1 is observable fact — published data from official UK sources. The baseline values for all 29 policy levers and 17 output metrics are drawn from these sources, cited in the baseline view.

Layer 2 is where modelling judgment enters. Each of the 71 edges in the causal graph represents a claim: “if X changes, Y changes in this direction, at this magnitude, with this delay.” These are informed by published research and policy analysis, but the specific magnitudes are author-assigned heuristics — not regression coefficients from a fitted model. This is by design: we trade econometric precision for inspectability and contestability.

Layer 3 handles the fact that some effects accumulate over time (a “stock” — e.g., trained nurses entering the workforce) while others operate at a rate (e.g., higher pay reducing vacancy rates immediately). It also encodes lag parameters — the delay between a policy change and its observable effect.

Layer 4 bundles lever configurations into named ideological starting points. These are not endorsements — they represent recognisable policy positions so users can explore trade-offs from a familiar starting point before adjusting individual levers.

3. The causal graph

The simulation engine is a weighted directed graph. Nodes are either policy levers (inputs the user controls) or output metrics (system states the model computes). Some nodes are both — they receive effects from upstream and propagate effects downstream. For example, nhs_staffing_pressure is influenced by spending and immigration policy, and in turn drives hospital waiting times and political risk.

Each edge carries four properties:

This diagram shows a simplified subset of the 71 edges in the model. The full edge catalogue is in section 9 below.

Critically, several nodes form feedback loops. Fiscal pressure constrains NHS and social care budgets, which worsens staffing, which increases hospital waiting times, which erodes public satisfaction, which increases political risk. The model propagates through these loops iteratively (3 passes with damping) to reach approximate equilibrium — not a mathematically proven fixed point, but a stabilised estimate.

4. Mathematical mechanics

The simulation proceeds in three steps: lever deviation, graph propagation, and score conversion.

5. Fiscal model

The budget model runs alongside — but partially independent of — the causal graph. It computes revenue, spending, and deficit from lever positions, then feeds the deficit back into the graph as fiscal pressure.

6. Time horizon and lag attenuation

Not all policies deliver results on the same timescale. Increasing social care worker pay might reduce vacancy rates within a year. Training new doctors takes five. The model handles this through two mechanisms:

7. Baseline data and sources

Every lever baseline and output starting value represents real UK data from 2024–2026. Sources include ONS, OBR, NHS England, DHSC, Home Office, Bank of England, IFS, Migration Observatory (Oxford), Resolution Foundation, House of Commons Library, and UK Parliament research briefings. Specific citations are displayed in the baseline view alongside each metric.

Baseline pressure scores are calibrated judgments. A score of 73 for NHS staffing pressure, for instance, reflects 100,165 vacancies, widespread delayed transfers of care, and a workforce plan that acknowledges a structural shortfall. These are not statistical derivations — they are informed starting points that anchor the relative movement when policies change.

8. Domain composite scoring

The report view aggregates individual metrics into five domain scores. Each domain is the unweighted arithmetic mean of its component metrics, with an inversion applied to metrics where higher values are beneficial (GDP, satisfaction) rather than harmful (pressure metrics):

9. Full edge catalogue

Every causal relationship in the model is listed below, grouped by source domain. This is the complete specification — there are no hidden edges. If a relationship isn't listed here, the model doesn't simulate it.

10. Known limitations and simplifications

No model this compact can represent the UK economy with precision. The following are deliberate simplifications we want you to know about — not bugs we hope you won't notice.

11. Challenge the model

If you believe a causal relationship is wrong, a magnitude is miscalibrated, a lag is unrealistic, or a significant pathway is missing entirely — we want to hear it. This model improves through challenge, not despite it.

Submitted challenges will be reviewed against the existing evidence base. Where a challenge leads to a model update, it will be documented in a public changelog with attribution. The goal is a model that gets less wrong over time, not one that claims to be right now.