# The Regulated to Fail Policy Simulator: downloadable model

Model 0.5.0. Generated from the simulator's data when the website was built (2026-10-14T15:50:51.828Z).

The simulator runs in the browser at https://regulatedtofail.com/simulator; how it works, and every source, is at https://regulatedtofail.com/simulator/method. These files let you check its assumptions, rerun its numbers or build on it. They are released under Creative Commons Attribution 4.0 (CC BY 4.0); the HESA, OfS, DfE and other source data keep their own licences.

## Files

- **simulator-model.json** — everything the model uses. Top-level keys:
  - `parameters`: 121 relationships and assumptions, keyed by id. Each has `kind` (accounting: a rule known exactly; empirical: measured from data; assumption: a judgement), `evidence` (strong, moderate, weak, none), `unit`, `central` and its plausible range `low`/`high` (or `byArchetype` for values that differ by type of provider), `sources` (each with what it supports) and a `note`.
  - `archetypes`: the OfS's seven provider groups, with their 2024-25 accounts (£m), TRAC cost recovery, the within-group distributions measured from HESA, regional mix and subject mix. The model builds one synthetic provider per real provider from these; no synthetic provider is a real one.
  - `policy`: current policy as dated schedules (`schedule`, then `thereafter`: hold, or index to CPI, RPI or earnings).
  - `exogenous`: inflation, earnings, Bank Rate, pay awards and the 18-year-old population.
  - `levers`, `leverDrawers`, `presets`, `shocks`: what a visitor can change.
  - `subjects`: the subjects tracked, with graduate earnings and employment (LEO) and how often each was withdrawn (HESA).
  - `calibrationTargets`: the published figures the model is checked against.
  - `behaviourEvidence`, `distributionProvenance`, `archetypeProvenance`: what the HESA and OfS data showed, and how it was used.
- **parameters.csv** — the parameters as a table, one row each.
- **current-policy.csv** — current policy's central run, year by year: the six gauges and the supporting figures (see the simulator for definitions), sector income and surplus (£m), and net liquidity days.
- **sensitivity.csv** — each parameter set to its low and then its high value, everything else central, and the effect on six outcomes in 2035-36, ranked by influence.

## Conventions

Money is £m unless stated. Academic years are labelled by the year they start (2030 is 2030-31). Indices are 100 in 2024-25. Fractions are written as decimals (0.03 is 3 per cent).

## The code

The engine is plain TypeScript (`src/lib/simulator/` in the website's repository). Results come from one synthetic sector (seed 272) and the central assumptions unless a file says otherwise.