What the model does
The simulator is a model of English higher education's finances from 2024–25, its base year, to 2035–36. It is illustrative: it lets you explore how specified policy changes propagate through the system, not what will happen. Its counts of providers in deficit, merging or closing are outputs of a model, not forecasts; it cannot say which providers will fail, what a policy costs the public finances in total, or whether lighter regulation is better overall (it counts regulation's costs, not its benefits). It runs in your browser; nothing you do is sent anywhere.
Each simulated year, in a fixed order:
- Policy, prices and people. The rules in force that year, inflation, pay awards, pension and National Insurance rates, and the number of 18-year-olds.
- Students. Each year's home undergraduate intake follows that year's 18-year-olds and each provider's reputation; taught and research postgraduates are recruited separately, so undergraduate policies do not move them. International entrants respond to visa rules, fees against a fee that keeps pace with prices (including the levy passed on) and reputation. Every cohort stays for the length of its course.
- Income and costs. Fees, grants and research income; staff and running costs, which move only partly with student numbers. Funding body grants are kept apart: the OfS teaching grant, Research England's QR and other grants each follow their own policy.
- Cash. The surplus or deficit, less capital spending, changes each provider's cash. A provider that ends a year owing more than its cash and current investments pays interest on the gap at Bank Rate plus a premium.
- Status. Comfortable, in deficit, in serious distress (cash below 10 per cent of a year's spending), or unable to pay its way.
- Responses. Providers under pressure freeze posts and capital spending, run severance schemes, push international recruitment, close courses, cut research, sell assets, and finally seek a merger or close. Providers in comfortable surplus rebuild and invest.
- Closures. A closed provider's continuing students finish their courses elsewhere if a place can be found — in its region first, then anywhere in England, each receiving provider taking at most a set share of its own students — or lose their place. Its future applicants mostly apply elsewhere, within the same limit.
- Consequences. Student experience, research standing and reputation follow those responses slowly — which is why a policy that looks fine in 2027 can look different in 2034.
Every relationship is one of three kinds, and each carries an evidence grade: accounting (a rule or identity, known exactly), measured (estimated from data) or assumption (a judgement). Of 121 parameters, 5 are accounting, 34 measured and 82 assumptions.
"Why did this happen?" on the simulator splits any difference between your policy and current policy into what each rule does on its own (re-running the model with just that rule changed) and what the rules do together.
The synthetic sector
The model has one synthetic provider for each of the 279 registered providers in the Office for Students' financial data, in the OfS's seven groups. No synthetic provider is a real one, and none is named after one. Each group's totals reproduce the OfS's 2024–25 accounts exactly; within a group, providers differ in size, margin, cash, international students, region and subjects as real providers in that group do, measured from HESA's provider-level data.
What the seven groups mean
The model uses the seven groups the Office for Students uses in its financial reports. The OfS sorts providers by asking, in order: are most of its students below degree level? Does it specialise in one or two subjects? How much teaching income (tuition fees and OfS grants) does it have? Further education colleges are not in the OfS's financial figures, so they are not in the model.
- Large research-intensive university (Research-intensive, 21 providers)
Teaching income (tuition fees and OfS grants) of more than £200m a year, making up less than 70% of all its income. Most of the rest comes from research grants and contracts.
The big research universities, most with 20,000 to 35,000 students. Almost all are members of the Russell Group.
- Large teaching-intensive university (Teaching-intensive, 14 providers)
Teaching income of more than £200m a year, making up more than 70% of all its income: it depends on tuition fees far more than on research.
Large universities, most with 22,000 to 32,000 students. Most are former polytechnics, which became universities in 1992.
- Medium-sized university (Medium, 43 providers)
Teaching income of £100m to £200m a year.
Mid-sized universities, most with 10,000 to 20,000 students: post-1992 universities alongside smaller research universities, several of them founded in the 1960s.
- Smaller university or college (Smaller, 53 providers)
Teaching income of less than £100m a year (or not known), for a provider that is neither a specialist nor teaching mainly below degree level.
A varied group: small universities (many of which began as teacher-training or church colleges), university colleges and private providers. Most have fewer than 10,000 students and many fewer than 1,000.
- Specialist institution (Specialist, 58 providers)
At least 75% of its students in one broad subject area, or at least 90% in two, where neither is creative arts and design. Specialism is tested before size, so a specialist can be any size.
Theological and Bible colleges, medical, veterinary and health schools, business and law schools, agricultural universities and some postgraduate research institutes. Many are very small; a few are large or research-heavy.
- Specialist creative institution (Creative, 45 providers)
A specialist (at least 75% of students in one broad subject area, or 90% in two) where creative arts and design is one of the two.
Conservatoires, drama and dance schools, art and design universities, film schools and popular music colleges. Mostly small, with costly studio and small-group teaching.
- Level 4 and 5 provider (Level 4-5, 45 providers)
More than half of its higher education students are studying at Level 4 or 5, below a bachelor's degree. Level 4 is the standard of a degree's first year (such as a Higher National Certificate, HNC); Level 5 of its second (a Higher National Diploma, HND, or a foundation degree); a bachelor's degree is Level 6. This is tested first, so such a provider is in this group whatever its size or subject.
Mostly private providers, most teaching business and management, almost all to UK students. Further education colleges teach most Level 4 and 5 courses, but their finances are overseen by the Department for Education, not the OfS, so they are not in the OfS's figures or in the model.
Source: Office for Students, Provider typologies 2022, and its provider data for the typical student numbers (full-time equivalent, 2020-21). The numbers of providers are those in the OfS's 2024-25 financial data.
Caveats
- The groups are the OfS's, not the six types in the design brief. The brief's split between global and civic research universities is represented by variation between synthetic providers within the large research-intensive group, not by a separate type.
- 2024-25 is audited for 247 providers; for 32 providers that had not yet submitted their 2025 return, the OfS used 2024-25 forecasts.
- OfS groups and TRAC peer groups are drawn on different criteria, and TRAC covers UK higher education institutions only. The TRAC figures attached to each group are borrowed, not measured for that group.
- Student numbers are full-time equivalents. The international student levy is charged per head, so the model converts with an assumed headcount-to-FTE ratio.
Subjects
Each provider teaches a mix of 25 subjects, from HESA's student data. Course closures fall on subjects in proportion to how often English providers withdrew them between 2019/20 and 2024/25. Graduate earnings and employment are the Department for Education's LEO figures five years after graduation; they move only with the mix of subjects taught. They are indices of what current students (at all levels and from every country) are studying, not projections of graduates entering England's workforce.
Subjects| Subject | Students, 2024/25 | Withdrawn by providers, 2019/20 to 2024/25 | Median earnings after 5 years | In work or study after 5 years |
|---|
| Medicine and dentistryTrains the public-service workforce | 137,000 | 3% | £56,803 | 92% |
|---|
| Nursing and midwiferyTrains the public-service workforce | 261,000 | 4% | £37,318 | 94% |
|---|
| Pharmacy, therapies and healthcare scienceTrains the public-service workforce | 158,000 | 6% | £37,078 | 91% |
|---|
| Other subjects allied to medicine | 179,000 | 6% | £36,812 | 90% |
|---|
| Teacher trainingTrains the public-service workforce | 81,000 | 6% | £34,400 | 92% |
|---|
| Education | 87,000 | 4% | £26,700 | 89% |
|---|
| Social workTrains the public-service workforce | 41,000 | 4% | £32,900 | 91% |
|---|
| Social sciences | 417,000 | 6% | £36,176 | 88% |
|---|
| Biological and sport sciences | 172,000 | 6% | £32,153 | 90% |
|---|
| Psychology | 231,000 | 6% | £29,554 | 89% |
|---|
| Veterinary sciences | 21,000 | 6% | £40,860 | 94% |
|---|
| Agriculture, food and related studies | 29,000 | 13% | £28,938 | 89% |
|---|
| Physical sciences | 105,000 | 6% | £39,011 | 90% |
|---|
| Mathematical sciences | 74,000 | 5% | £43,867 | 89% |
|---|
| Engineering and technology | 287,000 | 5% | £43,859 | 89% |
|---|
| Computing | 327,000 | 8% | £40,276 | 86% |
|---|
| Architecture, building and planning | 114,000 | 6% | £39,604 | 88% |
|---|
| Law | 248,000 | 1% | £33,300 | 88% |
|---|
| Business and management | 1,020,000 | 4% | £35,564 | 85% |
|---|
| Languages and area studies | 116,000 | 13% | £31,920 | 87% |
|---|
| History, philosophy and religion | 119,000 | 6% | £32,603 | 88% |
|---|
| Combined and general studies | 53,000 | 17% | £30,453 | 83% |
|---|
| Media, journalism and communications | 77,000 | 9% | £28,298 | 86% |
|---|
| Design, and creative and performing arts | 310,000 | 9% | £26,400 | 86% |
|---|
| Geography, earth and environmental studies | 68,000 | 13% | £35,323 | 90% |
|---|
Sources
- Financial sustainability of higher education providers in England 2026, Annex E: Sector financial data — Office for Students. Every 2024-25 income, expenditure, balance-sheet and student-number figure for each provider group, reproduced verbatim in raw/ofs-2026-annex-e-provider-groups.csv from the sheet 'All_data_output_peer_group'. Provider counts per group from the 'Notes' sheet.
- Financial sustainability of higher education providers in England 2026 (updated 15 June 2026) — Office for Students. Net liquidity days by provider group, 2024-25 (Table 17): lower quartile, mean and upper quartile.
- Annual TRAC 2024-25, Annex C: analysis by TRAC peer group — Office for Students. Activity cost shares and full economic cost recovery by TRAC peer group (Tables 1-3), reproduced in raw/trac-2024-25-peer-groups.csv.
- TRAC peer groups for 2024-25 (Annex 4.1b of the TRAC guidance) — TRAC Development Group. Peer group definitions: A medical school and research ≥20% of income; B other research ≥15%; C research 5-15%; D research <5% and income >£150m; E research <5% and income ≤£150m; F specialist music/arts.
- Provider typologies 2022 (December 2022 update) — Office for Students. How providers are sorted into the seven finance-typology groups, in order: more than half of higher education students at Level 4 or 5; specialist (at least 75% of students in one broad subject area or 90% in two, creative if one is creative arts and design); then qualifying income (OfS grants and tuition fees) over £200m, £100m-£200m or under £100m, with the largest split at 70% of total income. The typical sizes and kinds of provider in each group's 'typical' text are from the accompanying provider data (2020-21 student numbers).
- Financial sustainability of higher education providers in England 2026 (updated 15 June 2026): introduction — Office for Students. The OfS financial data covers higher education providers in England excluding further education colleges.
- HE Provider Data: Finance, tables 1, 3, 4, 5, 6, 7C and 12 (2015/16 to 2024/25) — HESA (Jisc). Provider-level income, expenditure, balance sheet, cash flow, tuition fees by domicile, research income by sponsor, funding body grants and staff costs by pension scheme. Last updated May 2026. Creative Commons Attribution 4.0.
- Provider typologies 2022: provider data — Office for Students. The finance typology group of each provider, by UKPRN.
- HE Provider Data: Finance, tables 1, 3, 4 and 12 (2015/16 to 2024/25) — HESA (Jisc). Provider-level income, expenditure (including restructuring costs and staff costs), cash, capital spending, staff FTE and region. Creative Commons Attribution 4.0.
- Table 49 - HE student enrolments by HE provider and subject of study 2014/15 to 2024/25 — HESA (Jisc). Enrolments by provider, CAH level 3 subject, level and mode of study. Last updated January 2026. Creative Commons Attribution 4.0.
- Graduate labour market outcomes (LEO), tax year 2023-24: CAH level 3 subject data — Department for Education. Median earnings and sustained employment or further study, five years after graduation, by CAH3 subject (first degree, all graduates). Rows used are kept verbatim in data/simulator/raw/leo-2023-24-cah3-first-degree-5yag.csv.
Every assumption
The value the model uses, the range a future uncertainty analysis will sample, the kind of relationship, the strength of the evidence, and the sources, with exactly what each source supports.
Executive pay and management
An illustrative executive payroll share, severance costs and experimental pay-setting rules. Salary distributions, median staff pay and progression responses are assumptions. Leadership quality, recruitment, retention, morale and management capacity effects are not quantified.
Executive salaries as a share of total salaries
1% of salaries
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0.5% of salaries to 2% of salaries
Illustrative 1% executive salary pool, not a measured sector estimate. Applied to each provider’s salary bill after ordinary pay growth, staffing/research changes and merger savings. Covers the vice-chancellor and other senior executives, not all professional services staff. The model has no executive payroll dataset; the same share is assumed across all seven provider groups. Employer on-costs use the provider group’s blended rate and change proportionally. Fixed benefits, pension caps and individual NI thresholds are not separately modelled. Evidence supports caution about wider performance effects, not this cost-base value. Set this share to zero to remove all direct executive-policy costs and savings.
Parameter executive.salaryShare
Executive severance cost per annual salary cost removed
0.75 multiple
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0 multiple to 1.5 multiple
Illustrative one-off transition cost when executive headcount falls: 0.75 times the annual employer cost of the posts removed, before any simultaneous change in salary rates. Not a measured estimate for executives and not a legal entitlement. Charged only on the reduction from the preceding year’s headcount setting; salary reductions alone incur no severance. Management-capacity effects, recruitment expenses for added posts, contractual delays and fixed benefits are not quantified.
No source: a judgement, open to challenge.
Parameter executive.severanceMultiple
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 150,000 £ per executive a year to 300,000 £ per executive a year
Illustrative arithmetic mean of executive salaries in 2024–25, not a measured sector figure. Used with a lognormal distribution only to estimate the share of payroll affected by salary caps and ratios; the total executive pool still comes from executive.salaryShare. Grows with ordinary pay, the remuneration control and experimental progression. The same hypothetical distribution is used in all provider groups; individual salaries, posts and benefits are not observed.
No source: a judgement, open to challenge.
Parameter executive.meanSalary
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0.15 standard deviation of log salary to 0.6 standard deviation of log salary
Illustrative dispersion of individual executive salaries around the assumed arithmetic mean. A lognormal distribution ensures positive salaries and permits a cap to affect only the upper tail. Not fitted to observed executive pay. Zero gives identical salaries. Constraints are evaluated afresh against the unconstrained salary path each year, with no contractual delay or pay stickiness.
No source: a judgement, open to challenge.
Parameter executive.salaryLogSd
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 30,000 £ per staff member a year to 45,000 £ per staff member a year
Illustrative 2024–25 median salary for the denominator of the executive/staff salary ratio, not an observed median or the HESA mean salary. It follows the ordinary pay index, independently of executive policy; changing executive headcount does not change this median. The ratio and statutory-style ceiling both constrain salary, with proportional employer on-costs, not separately assessed total benefits.
No source: a judgement, open to challenge.
Parameter executive.medianStaffSalary
Annual executive pay ratchet removed by symmetric benchmarking
0.5% a year
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0% a year to 2% a year
Hypothetical annual premium treated as embedded in the current executive pay path. Symmetric benchmarking divides that path by (1 + premium) for each year the rule operates. This is a reduced-form counterfactual, not an estimated causal coefficient, an explicit peer-network simulation or a replication of Gmeiner et al. The current pay path is unchanged when this experiment is off. Accumulated progression differences persist if a later schedule changes the rule.
Parameter executive.benchmarkPremium
Maximum annual performance-linked executive pay adjustment
2% of salary
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0% of salary to 5% of salary
Hypothetical annual progression rule: take the previous year’s mean experience and research-standing index relative to 2024–25, subtract one, then clip to plus/minus this limit. Multiply the executive pay path by one plus this adjustment. A score of 0.99 gives a 1% reduction relative to ordinary pay growth; scores of 0.98 or below reach the central 2% limit. Scores above one permit increases. This is a salary response to past model outcomes, not a claim that higher pay causes better outcomes. No direct leadership-quality, recruitment, retention or morale response is modelled.
No source: a judgement, open to challenge.
Parameter executive.performanceAwardLimit
Money: costs, income and the public purse
How pay, prices, fees, research income and the government ledger move.
- Kind
- Accounting: a rule or identity, known exactly
- Evidence
- Strong
- Checked
- Yes, 2026-10-03
The rate at which research council grants in providers' 2024-25 accounts were paid. Research council income in later years scales with the UKRI rate in force (the cost-recovery lever) relative to this, while the volume of research, and so its cost, does not.
Parameter research.ukri.fecShare
Real growth in international fees
in line with CPI
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- CPI − 1 points a year to CPI + 2 points a year
The baseline keeps international fees constant in real terms, as the DfE's levy impact analysis does. The upper end reflects providers' own forecasts.
Parameter income.intlFeeGrowthMinusCpi
Real growth in postgraduate home fees, contracts and other income
in line with CPI
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- CPI − 1 points a year to CPI + 1 points a year
Covers unregulated home postgraduate fees, education contracts, accommodation and other commercial income, investment income and donations: all assumed to keep pace with prices.
No source: a judgement, open to challenge.
Parameter income.otherGrowthMinusCpi
Real growth in research grant and contract income
CPI − 0.5 points a year
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- CPI − 2 points a year to CPI + 1 points a year
Research councils are about a fifth to a quarter of research income; charities, government departments and industry make up the rest. Flat-cash UKRI budgets pull the total slightly below inflation.
Parameter research.grantGrowthMinusCpi
Future national pay awards relative to inflation
in line with CPI
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- CPI − 1 points a year to CPI + 0.5 points a year
Applied from 2027-28, after the known 1.4 per cent (2025-26) and 2 per cent (2026-27) awards. Recent awards have been at or below CPI because institutions could not afford more; awards matching CPI is the neutral assumption.
No source: a judgement, open to challenge.
Parameter costs.payAwardMinusCpi
Pay drift (incremental progression net of staff turnover)
0.5% a year
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 0% a year to 1.5% a year
Staff move up the pay spine each year; leavers are replaced lower down. The net effect adds to the pay bill on top of the national award even with unchanged staff numbers.
No source: a judgement, open to challenge.
Parameter costs.payDrift
Real growth in non-pay costs
in line with CPI
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- CPI − 1 points a year to CPI + 1.5 points a year
Applies to other operating expenses and depreciation. Energy, insurance, software licences and maintenance have recently risen faster than CPI; the upper end reflects that continuing.
No source: a judgement, open to challenge.
Parameter costs.nonPayGrowthMinusCpi
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- By type of provider
- Research-intensive 1.351 ratio; Teaching-intensive 1.006 ratio; Medium 1.181 ratio; Smaller 1.47 ratio; Specialist 1.497 ratio; Creative 2.249 ratio; Level 4-5 1.374 ratio
Each group's measured 2024-25 ratio is its providers' normal capital spending. A provider under pressure spends less (behaviour.capexCutLevel), one in comfortable surplus more (behaviour.investLevel), both measured against the median provider, so across the sector the three average out near the measured ratio. Spending above depreciation was partly financed by capital grants (already in income) and borrowing; the model holds borrowing constant, so the high end of the range assumes some of it is debt-financed.
Parameter finance.capexToDepreciation
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 0.015 annual rate above Bank Rate to 0.05 annual rate above Bank Rate
A provider whose borrowing exceeds its cash and current investments at the year end must have borrowed to cover the gap (a revolving credit facility, overdraft or deferred payments) and pays interest on it the following year. Lending to distressed providers is priced case by case; three points over Bank Rate is an illustrative premium for an unsecured facility.
No source: a judgement, open to challenge.
Parameter finance.shortTermPremium
Share of student loan outlay not expected to be repaid (RAB charge)
33% of outlay
- Kind
- Measured: estimated from data
- Evidence
- Strong
- Plausible range
- 25% of outlay to 40% of outlay
- Checked
- Yes, 2026-10-03
Used to turn fee income paid through loans into a cost to the taxpayer. The average charge is applied to all fee income; the charge on an extra pound of fee is probably higher, since more of it is written off.
- Student loan forecasts for England 2025-26 — Department for Education. 33% of Plan 5 full-time higher education loan outlay in 2025-26 is not expected to be repaid; forecast 34% in 2029-30. Released 9 July 2026.
Parameter government.rabCharge
Share of home undergraduate fee income paid through student loans
85% of fee income
- Kind
- Accounting: a rule or identity, known exactly
- Evidence
- Strong
- Checked
- Yes, 2026-10-03
Calculated from the provider-level HESA file (data/simulator/provider-level).
- HE Provider Data: Finance 2024/25, Table 6 — HESA (Jisc). English providers' UK-domiciled undergraduate fee income in 2024/25: £9.5bn from SLC and the devolved equivalents out of £11.2bn (85%); the rest is paid by employers, the NHS, sponsors and students directly.
Parameter government.homeFeeLoanShare
Direct costs that move with student numbers
20% of fee income
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 10% of fee income to 40% of fee income
Teaching materials, hourly-paid teaching, student services and some estates costs follow students; most costs do not, and change only when providers decide to cut (the decision ladder). Applied to the difference between fee income and what it would be at 2024-25 student numbers.
Parameter costs.variableShareOfFees
Response of QR funding to research standing at the REF
1 elasticity
- Kind
- Assumption: a judgement
- Evidence
- Moderate
- Plausible range
- 0.5 elasticity to 1.5 elasticity
QR is shared by quality-weighted research volume, so a provider's share moves roughly in proportion to its research standing relative to everyone else's. The total is set by the QR lever.
- REF 2029 — UK higher education funding bodies. REF outcomes inform the allocation of about £2bn a year of research funding; submissions in autumn 2028, results in December 2029.
Parameter research.qrQualityElasticity
First academic year funded on REF 2029 results
2030–31
- Kind
- Accounting: a rule or identity, known exactly
- Evidence
- Strong
Results in December 2029 inform allocations from 2030-31, as REF 2021 results (May 2022) did from 2022-23. Research standing is read in 2028-29, the year of submission.
- REF 2029 — UK higher education funding bodies. Results December 2029.
Parameter research.refFundingFromYear
Cost of the REF to institutions
£65m a year
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- £45m a year to £90m a year
£454m over REF 2021's seven-year cycle. Shared among providers in proportion to their research spending. A light-touch REF halves it; an annual one doubles it; abolition removes it.
Parameter research.refCostPerYear
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 35% of home undergraduates to 50% of home undergraduates
The maintenance support control is the extra grant for each eligible student; its cost is that amount times this share of home undergraduates. Proposals that name a coverage (the Liberal Democrats' 35 per cent, rising to 50) are close to this range.
Parameter government.maintenanceEligibleShare
Students
Home and international recruitment, and what moves it.
International students per international FTE
1.05 headcount per FTE
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 1 headcount per FTE to 1.15 headcount per FTE
The levy is charged per head; OfS data are in FTE. International students are overwhelmingly full-time because of visa conditions, so headcount is close to FTE. Students who start mid-year and continue beyond the anniversary are charged in two levy years, which pushes the ratio above 1.
No source: a judgement, open to challenge.
Parameter students.intlHeadcountPerFte
Share of international students exempt from the levy
5% of headcount
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 0% of headcount to 10% of headcount
The exempt categories are not yet final. The worked example is illustrative, not an estimate.
Parameter levy.exemptShare
Response of home undergraduate numbers to the 18-year-old population
1 elasticity
- Kind
- Assumption: a judgement
- Evidence
- Moderate
- Plausible range
- 0.5 elasticity to 1.2 elasticity
1.0 means participation rates stay constant, so student numbers move with the cohort. The undergraduate population responds to the average of the last three entry cohorts. From Phase 3, demand also responds to policy and reputation, and recruitment shifts between groups.
Parameter students.homeDemographyElasticity
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- -0.8 % change in entrants per % change in fees to -0.3 % change in entrants per % change in fees
The immediate response of new non-EU undergraduate entrants to a fee rise. Public First's study was commissioned by a group of universities opposing the levy; the DfE used the same estimates in its own analysis.
Parameter intl.priceElasticity.ugShortRun
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- -0.45 % change in entrants per % change in fees to 0 % change in entrants per % change in fees
The response once recruitment has adjusted: smaller than the short-run effect, as some lost demand returns.
Parameter intl.priceElasticity.ugLongRun
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- -1.1 % change in entrants per % change in fees to -0.5 % change in entrants per % change in fees
Postgraduates, mostly on one-year taught courses, are more price-sensitive than undergraduates. Applied to research students as well, for want of a separate estimate.
Parameter intl.priceElasticity.pgShortRun
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- -0.7 % change in entrants per % change in fees to -0.2 % change in entrants per % change in fees
As for the undergraduate long-run elasticity.
Parameter intl.priceElasticity.pgLongRun
Years for half the short-run fall in demand to recover
5 years
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- 3 years to 8 years
Demand moves to the short-run response when fees change and drifts towards the long-run response, closing half the gap every this-many years.
Parameter intl.priceAdjustmentHalfLife
- Kind
- Assumption: a judgement
- Evidence
- Moderate
- Plausible range
- 50% of the levy per student to 100% of the levy per student
The DfE's estimates (an income loss of £270-330m against receipts of £445-480m) imply providers recover most of the levy through fees; Public First assumed slightly more than full pass-through. Providers are assumed to raise fees for all international students, not only those above the allowance.
Parameter levy.passThrough
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- -0.25 proportional change in taught entrants to -0.08 proportional change in taught entrants
About a quarter of Graduate visa holders (73% × 34%) say they would not have come without the route, and about a third of students use it, which gives roughly −8%; prospective students who never use the route also weigh it, which raises the effect; −25% applies the survey response to all students. Shorter routes are interpolated in proportion to the months removed from 24, so 18 months is about −4%. Applies to undergraduate and taught postgraduate entrants; doctoral graduates keep three years.
- Rapid Review of the Graduate Route (May 2024) — Migration Advisory Committee. In a Home Office survey of Graduate visa holders, 73% of those aware of the route said it influenced their choice; of these, 34% would not have come to the UK without it. About 32% of international students go on to the route. In the QS survey, 30% of prospective students were 'much more' and 29% 'somewhat more' likely to consider the UK if the route were extended to three years.
Parameter intl.graduateRoute.abolitionEffect
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 0 proportional change in taught entrants to 0.1 proportional change in taught entrants
Stated intentions in the QS survey overstate behaviour; the central value takes a small fraction of the 59% who said they would be more likely to consider the UK.
- Rapid Review of the Graduate Route (May 2024) — Migration Advisory Committee. In a Home Office survey of Graduate visa holders, 73% of those aware of the route said it influenced their choice; of these, 34% would not have come to the UK without it. About 32% of international students go on to the route. In the QS survey, 30% of prospective students were 'much more' and 29% 'somewhat more' likely to consider the UK if the route were extended to three years.
Parameter intl.graduateRoute.extensionEffect
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 0.03 proportional change in taught postgraduate entrants to 0.15 proportional change in taught postgraduate entrants
Main-applicant applications fell 14% in 2024 after the ban, but the MAC notes other causes (the naira's collapse, perceptions of hostility) and that 'the precise magnitude is currently hard to discern'. The central value attributes a little over half of that fall, on taught postgraduates only, to the ban.
- Rapid Review of the Graduate Route (May 2024) — Migration Advisory Committee. In a Home Office survey of Graduate visa holders, 73% of those aware of the route said it influenced their choice; of these, 34% would not have come to the UK without it. About 32% of international students go on to the route. In the QS survey, 30% of prospective students were 'much more' and 29% 'somewhat more' likely to consider the UK if the route were extended to three years.
- International applications to the UK dip in 2024 — The PIE News. Home Office data: main-applicant study visa applications fell 14% in 2024 and dependant applications 84%, after the January 2024 dependants restriction.
Parameter intl.dependantsEffectPgt
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- -0.1 % change in entrants per % change in real fee to 0 % change in entrants per % change in real fee
- Checked
- Yes, 2026-10-03 — Bibliographic details and headline finding confirmed; the central value of zero is a judgement from it.
Because fees are paid through income-contingent loans, the headline fee has little effect on whether people go to university. The low end allows a small effect.
Parameter home.feeElasticity
Length of an undergraduate course
3 years
- Kind
- Accounting: a rule or identity, known exactly
- Evidence
- Strong
Undergraduates on roll are the last three years' entrants, so a change in entrants takes three years to reach full effect on income.
Parameter students.ugCourseYears
Length of a research degree
4 years
- Kind
- Assumption: a judgement
- Evidence
- Moderate
- Plausible range
- 3 years to 5 years
Full-time PhDs are funded for three to four years and often take longer.
No source: a judgement, open to challenge.
Parameter students.pgrCourseYears
Response of home recruitment to reputation
0.5 elasticity
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 0.2 elasticity to 1 elasticity
Applications respond to reputation, but modestly. The model's reputation index is not a league-table rank, so the elasticity is a judgement informed by the study, not an estimate from it.
Parameter demand.homeReputationElasticity
Response of international recruitment to reputation
1 elasticity
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 0.5 elasticity to 2 elasticity
International applicants rely more on rankings than home applicants, so the response is taken as twice as large. This is the loop by which research cuts eventually cost international students.
No source: a judgement, open to challenge.
Parameter demand.intlReputationElasticity
Effect of visa costs and friction on international entrants
5% of entrants per step
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 2% of entrants per step to 10% of entrants per step
A step from 'as now' to 'higher' (or 'lower') changes international entrants by this fraction. Visa and health surcharge costs add several thousand pounds to a degree; the effect is scaled to be smaller than the levy's, which is passed on in fees.
No source: a judgement, open to challenge.
Parameter intl.visaFrictionEffect
How sharply a targeted cut in places falls on low-earning subjects
2 elasticity
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 1 elasticity to 3 elasticity
When a cut in home undergraduate places is aimed at courses with low graduate earnings, each subject's share of the cut is scaled by (average graduate earnings / its graduates' earnings) to this power, with no subject losing more than 95 per cent of its entrants. Parties proposing targeted cuts have not said how courses would be chosen; this is one reading, using the same earnings data (LEO) as the rest of the model.
No source: a judgement, open to challenge.
Parameter places.earningsTargeting
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- 2% of home entrants per £1,000 to 6% of home entrants per £1,000
3.95 percentage points of participation per £1,000 is about a tenth more entrants among students eligible for grants; with roughly two in five students eligible, about 4 per cent overall.
Parameter home.maintenanceEffect
How providers respond
The decision ladder: when providers act, and what each response costs and saves. Most sizes are measured from what English providers did after deficits in 2015/16-2024/25 (HESA).
Consecutive deficit years before a provider starts restructuring
2 years
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 1 years to 3 years
The design brief's two consecutive deficits. Restructuring becomes markedly more common after the second deficit year in HESA's data. A provider whose cash falls below the low-liquidity threshold while in deficit acts after one.
Parameter behaviour.deficitYearsBeforeCuts
Liquidity below which a provider starts saving cash, even in surplus
90 days of spending
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 60 days of spending to 120 days of spending
Providers with under about 120 days of cash spend visibly less on capital than those with more; the restraint is similar from 30 to 120 days. 90 days sits inside that range; it is the trigger for the first rung (recruitment freeze and capital cut).
Parameter behaviour.cautionLiquidityDays
Staff lost each year to a recruitment freeze
2% of staff a year
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 1% of staff a year to 3% of staff a year
Providers in deficit grew staff numbers about two percentage points a year more slowly than comparable providers in modest surplus. The model has no underlying growth in staff, so the freeze is entered as a fall of that size: posts left unfilled as people leave.
Parameter behaviour.freezeAttrition
Staff leaving in one voluntary severance programme
3% of staff
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 2% of staff to 5% of staff
A median programme costing 2.0 per cent of staff costs, at three-quarters of a year's pay per leaver (behaviour.severanceCostMultiple), buys about 2.7 per cent of staff; 3 per cent is that, rounded. Schemes are open to all staff, so research activity falls with teaching staff at research-active providers. Savings start the year after the programme.
Parameter behaviour.severanceShare
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 0.5 years of the leaver's staff cost to 1.25 years of the leaver's staff cost
Voluntary schemes pay statutory redundancy plus an enhancement, typically several months' pay; pension strain costs can add to it. No sector-wide figure is published.
Parameter behaviour.severanceCostMultiple
Years between restructuring programmes at a provider still in deficit
3 years
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 2 years to 4 years
Roughly one programme every three years for a provider that stays in deficit. The model does not reproduce the 18 per cent background rate of restructuring among providers in surplus: reorganisations that do not respond to deficits are left in the base-year cost lines.
Parameter behaviour.restructuringIntervalYears
Capital spending of a provider under pressure, relative to normal
73%
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- 55% to 90%
Providers in deficit spend 0.73 times what the median provider spends on capital, relative to depreciation. The shortfall against normal spending accumulates as a maintenance backlog, which weighs on student experience until it is made up.
Parameter behaviour.capexCutLevel
Capital spending of a provider in comfortable surplus, relative to normal
124%
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- 110% to 140%
Providers with surpluses above behaviour.recoveryMargin and cash above the caution level invest more than depreciation, as they do in HESA's data. Without this, surpluses would pile up as cash that real providers spend on their estates.
Parameter behaviour.investLevel
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0% of normal capital spending to 50% of normal capital spending
A recovered provider spends a quarter more than normal on its estate until the deferred work is done.
No source: a judgement, open to challenge.
Parameter behaviour.capexCatchUp
Surplus at which a provider starts rebuilding
3% of income
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 1% of income to 5% of income
Above a 3 per cent surplus (with liquidity above the caution level) providers hire back towards their 2024-25 student-staff ratio and restore research activity.
Parameter behaviour.recoveryMargin
Fastest rebuilding of staff or research a year
3% a year
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 2% a year to 5% a year
Recovery is slower than decline: hiring takes time and providers that have cut stay cautious.
Parameter behaviour.hiringRate
Surplus at which a provider expands beyond its 2024-25 staffing
6% of income
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 4% of income to 8% of income
Providers with large surpluses spend them: staff grow by behaviour.hiringRate a year even beyond the 2024-25 student-staff ratio, which improves student experience and closes the surplus. This, the recruitment freeze and severance are the model's routes by which margins revert towards their group, as HESA shows they do.
Parameter behaviour.expansionMargin
Extra international entrants from one recruitment push
5% of entrants
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0% of entrants to 10% of entrants
More agents, new markets and partnerships. The effect on the UK total is smaller (behaviour.intlNewToUkShare): much of it is students who would have gone to another English provider.
No source: a judgement, open to challenge.
Parameter behaviour.intlPushEffect
Most a provider can add to international recruitment by effort
10% of entrants
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 5% of entrants to 20% of entrants
Two pushes at most: demand, visas and reputation, not effort, set the scale of international recruitment.
No source: a judgement, open to challenge.
Parameter behaviour.intlPushMax
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 10% of the first year's fee to 30% of the first year's fee
Journalistic evidence from freedom of information requests; no official series exists.
Parameter behaviour.agentCommission
Share of extra international recruits who would not otherwise have come to England
50%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 20% to 80%
The rest are taken from other English providers, so if every provider pushes, the sector gains little. Taken from other providers in proportion to their international entrants, the following year.
No source: a judgement, open to challenge.
Parameter behaviour.intlNewToUkShare
Staff cost removed by one round of course closures
4% of teaching staff
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 2% of teaching staff to 6% of teaching staff
Closures target low-enrolment courses, which cost more per student than average, so a round removes more staff than students. Savings start two years later, after teach-out.
No source: a judgement, open to challenge.
Parameter behaviour.closureStaffShare
Students lost by one round of course closures
1.5% of students
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0.5% of students to 3% of students
Low-enrolment courses carry few students, and some of those who would have taken them choose another course at the same provider.
No source: a judgement, open to challenge.
Parameter behaviour.closureStudentShare
Course provision lost by one round of closures
6% of provision
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 3% of provision to 10% of provision
Subject breadth falls by more than student numbers: the courses that close are small ones.
No source: a judgement, open to challenge.
Parameter behaviour.closureBreadthShare
Research activity cut in one round
10% of research activity
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 5% of research activity to 20% of research activity
Research recovers about 70 per cent of its full cost at research-intensive providers (TRAC), so cutting it saves money at once; the QR income that depends on it falls only after the next REF. Applied to providers whose research is at least 10 per cent of their costs.
No source: a judgement, open to challenge.
Parameter behaviour.researchCutShare
Compulsory staff reduction in a liquidity crisis
5% of teaching staff
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 3% of teaching staff to 8% of teaching staff
The step after voluntary schemes: compulsory redundancies, larger classes and fewer optional modules. Costs the same per leaver as severance.
No source: a judgement, open to challenge.
Parameter behaviour.ssrRiseShare
Cash raised by selling assets
5% of annual income
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 2% of annual income to 10% of annual income
A one-off disposal of land, buildings or investments. Gains on disposal are excluded from the surplus the model reports (as in the OfS adjusted surplus), so the sale affects cash only.
No source: a judgement, open to challenge.
Parameter behaviour.assetSaleShare
Share of providers with assets they could sell
50% of providers
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 25% of providers to 75% of providers
Drawn once per provider with the population.
Parameter behaviour.assetSaleAvailability
Slow-moving consequences
Student experience, research standing and reputation, which follow resources with a lag, and what steers which courses close.
Time for student experience to respond to resources
2 years
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 1 years to 4 years
The design brief's two to three years for staff cuts to show up in student experience: larger classes are felt at once, survey results and continuation a year or two later.
No source: a judgement, open to challenge.
Parameter stocks.experienceTimeConstant
Time for research standing to respond to research activity
4 years
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 3 years to 7 years
Outputs lag funding and people by years; standing is what the next REF measures.
No source: a judgement, open to challenge.
Parameter stocks.researchTimeConstant
Time for reputation to respond to experience and research
5 years
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 3 years to 8 years
League tables use rolling data and applicants' views change slowly.
No source: a judgement, open to challenge.
Parameter stocks.reputationTimeConstant
Response of student experience to the student-staff ratio
0.5 elasticity
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0.2 elasticity to 1 elasticity
A 10 per cent rise in students per member of staff lowers the experience index by about 5 per cent. There is no robust causal estimate for English universities; this is the model's most important unmeasured link.
No source: a judgement, open to challenge.
Parameter experience.ssrElasticity
Response of student experience to the condition of the estate
0.3 elasticity
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0.1 elasticity to 0.5 elasticity
Estate condition is 1 less the maintenance backlog as a share of five years of normal capital spending.
Parameter experience.estateElasticity
How much an emphasis on graduate earnings steers closures
2 elasticity
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 1 elasticity to 3 elasticity
With a strong emphasis on earnings, a subject's chance of closure is scaled by (average graduate earnings / its graduates' earnings) to this power, so arts and humanities go first. The brief's point: what is measured changes what institutions do.
No source: a judgement, open to challenge.
Parameter earnings.emphasisElasticity
Distress, mergers and closure
When a provider is in trouble, when it merges, when it closes — and what a restructuring regime would change.
- Kind
- Accounting: a rule or identity, known exactly
- Evidence
- Strong
- Checked
- Yes, 2026-10-03
A definition, not an estimate: the OfS's own threshold, about 37 days of spending. Providers below it are shown red.
Parameter status.lowLiquidityShare
Net liquidity at which a provider can no longer pay its way
-30 days of spending
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- -60 days of spending to 0 days of spending
Net liquidity (cash and current investments less overdrafts) can fall below zero while a provider draws on revolving credit or delays payments: five synthetic providers, like a handful of real ones, start 2024-25 below zero. Beyond about a month of spending owed, the model treats the provider as unable to continue without a merger or exit. There is no published threshold: lenders, the OfS and boards act case by case.
No source: a judgement, open to challenge.
Parameter status.insolvencyDays
Liquidity a provider needs to take over another
120 days of spending
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 90 days of spending to 180 days of spending
A partner must be financially comfortable (in surplus, liquidity above this level) and keep at least half this level after absorbing the other provider's cash shortfall. Partners in the same region are preferred, then the same type of provider elsewhere.
No source: a judgement, open to challenge.
Parameter merger.partnerMinDays
Chance each year that a provider in trouble finds a willing partner
50% chance a year
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 25% chance a year to 75% chance a year
Even where a financially able partner exists, a merger needs a strategic fit, due diligence, lender and OfS consent and agreement between governing bodies; many discussions fail. Among able partners in the region, one is chosen with a chance proportional to its income; a provider that has absorbed another in the last merger.integrationYears is not available.
Parameter merger.partnerChance
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 2% of the absorbed provider's annual spending to 10% of the absorbed provider's annual spending
Systems, rebranding, legal costs and severance, in the year after the merger.
No source: a judgement, open to challenge.
Parameter merger.integrationCost
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 2% of the absorbed provider's staff and other costs to 10% of the absorbed provider's staff and other costs
Duplicated back-office functions; reached over merger.integrationYears.
Parameter merger.integrationSaving
Years for a merger to reach its full saving
3 years
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 2 years to 5 years
Savings phase in evenly.
No source: a judgement, open to challenge.
Parameter merger.integrationYears
Years a provider can remain unable to pay its way before it closes
2 years
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 1 years to 3 years
If no partner can be found, a provider below the insolvency threshold for this many consecutive years closes. England has no special administration regime for universities; how long lenders, the OfS and government would let a provider continue is untested.
No source: a judgement, open to challenge.
Parameter exit.yearsInsolvent
Continuing home students of a closed provider who finish their courses elsewhere
50%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 25% to 75%
Student protection plans aim to move continuing students. Those placed go to providers in the closed provider's region first, then elsewhere in England, within each receiving provider's capacity (exit.receiverCapacity), and stay only until they would have finished. The rest lose their place. Future applicants are a separate assumption (exit.applicantShare).
No source: a judgement, open to challenge.
Parameter exit.homeTransferShare
Continuing international students of a closed provider who finish their courses at another English provider
25%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0% to 50%
International students are mobile but sensitive to disruption, and some return home or transfer abroad. Those placed stay until they would have finished. Closures may also harm the UK's reputation, which the model does not include.
No source: a judgement, open to challenge.
Parameter exit.intlTransferShare
Home applicants who would have gone to a closed provider and enrol elsewhere instead
75%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 50% to 90%
Demand for places follows the number of 18-year-olds, not the number of providers, so most of a closed provider's future applicants apply elsewhere, in its region first. The rest do not enter higher education, which is what the regional access gauge records. Placed within receiving providers' capacity (exit.receiverCapacity).
No source: a judgement, open to challenge.
Parameter exit.applicantShare
International applicants who would have gone to a closed provider and choose another English provider
50%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 25% to 75%
International applicants choose between countries as well as providers, so fewer stay in England than home applicants.
No source: a judgement, open to challenge.
Parameter exit.intlApplicantShare
Most extra students a provider can take from closed providers
20% of its own students
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 10% of its own students to 35% of its own students
A limit on teaching space, staff and accommodation. Continuing students taken over in any year are at most this share of the receiving provider's own students; its extra entrants from closed providers' applicants, over all closures together, at most this share of its own entrants. Beyond it, students are placed further away or lost. Without a limit, a small provider surrounded by closures could absorb several times its own size.
No source: a judgement, open to challenge.
Parameter exit.receiverCapacity
How much earlier preventative monitoring makes providers act
30 days of spending
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 15 days of spending to 60 days of spending
Preventative monitoring raises the cash level at which providers start saving by this much, and makes providers in serious distress look for partners before their cash runs out; reactive monitoring lowers it as much.
No source: a judgement, open to challenge.
Parameter monitoring.cautionDaysShift
Effect of the merger regime on the chance of a merger
50%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 25% to 75%
A facilitated regime raises the chance of finding a willing partner by this fraction, and cuts merger costs by as much; a difficult one does the reverse.
No source: a judgement, open to challenge.
Parameter merger.regimeEffect
Restructuring regime: improvement in the chance of a merger
1.5 multiple
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 1.2 multiple to 2 multiple
Under a formal regime someone convenes lenders, the governing body and receiving institutions early, so more distressed providers find a partner — starting when cash is low after repeated deficits, before it runs out.
Parameter resolution.partnerChanceFactor
Restructuring regime: home students moved when a provider closes
85%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 70% to 95%
Receiving capacity is arranged in advance, so most continuing students are placed (still within each receiving provider's capacity). Without a regime the model assumes half.
No source: a judgement, open to challenge.
Parameter resolution.homeTransferShare
Restructuring regime: international students moved when a provider closes
50%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 30% to 70%
Against a quarter without a regime.
No source: a judgement, open to challenge.
Parameter resolution.intlTransferShare
Restructuring regime: public cost of each student moved
3,000 £ per student
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 1,500 £ per student to 6,000 £ per student
Temporary funding for teach-out and transfer, borne by the taxpayer (the book leaves the source open: lending, a levy or insurance).
No source: a judgement, open to challenge.
Parameter resolution.costPerStudent
Restructuring regime: extra cost of borrowing
0.3% of borrowing a year
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0% of borrowing a year to 1% of borrowing a year
Government's objection that a regime might unsettle lenders, entered as a quarter of a point on every provider's interest.
Parameter resolution.lenderPremium
Regulation and collaboration
The costs of compliance and the savings from shared services.
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 15% of expenditure (excluding depreciation and interest) to 30% of expenditure (excluding depreciation and interest)
Back-office administration, IT, procurement, payroll and estates management. HESA expenditure by activity (Table 8) would measure this directly; not yet imported.
No source: a judgement, open to challenge.
Parameter services.addressableShare
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- 6% of the function's cost to 10% of the function's cost
Applied to the share of addressable spending moved into shared services.
Parameter services.savingRate
Share of addressable spending shared within five years if actively facilitated
40%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 20% to 60%
Existing shared services reach only a subset of universities; facilitation (convening, start-up funding, a catalogue) widens take-up.
No source: a judgement, open to challenge.
Parameter services.adoptionFacilitated
Share shared within five years if facilitated and VAT is neutralised
70%
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 50% to 90%
Removing the VAT penalty makes most shared arrangements worthwhile, so take-up is higher.
No source: a judgement, open to challenge.
Parameter services.adoptionVatNeutral
Share of the saving lost to VAT when it is not neutralised
50% of gross saving
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 25% of gross saving to 100% of gross saving
Services one university supplies to another attract 20 per cent VAT, which the buyer cannot recover, unless they qualify for the cost-sharing exemption, whose conditions are strict. A 6-10 per cent saving cannot absorb a 20 per cent charge, so only arrangements that qualify go ahead, and they cost more to run.
Parameter services.vatDrag
Years to reach full adoption of shared services
5 years
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 3 years to 8 years
Procurement cycles, system migrations and staff consultation make this slow.
No source: a judgement, open to challenge.
Parameter services.rampYears
Set-up cost of shared services, as a multiple of the eventual annual saving
1 multiple
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0.5 multiple to 2 multiple
Spent over the first two years, so the policy costs money before it saves any.
No source: a judgement, open to challenge.
Parameter services.setupCostMultiple
Regulatory compliance cost of a provider with £100m income
£0.8m a year
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- £0.5m a year to £1.2m a year
Moorhouse's 17.6 compliance FTE at a typical university, plus OfS fees and systems. Scaled to other sizes by regulation.scaleElasticity; with the central values the 279 providers' compliance costs total about £0.2bn, consistent with Moorhouse's £150-200m for UUK members once smaller providers are added.
- Understanding the burden of regulation — Moorhouse Consulting for Universities UK. An average of 17.6 FTE per university dedicated solely to regulatory compliance; of the order of £150-200m a year across 116 English universities.
Parameter regulation.costAtReference
- Kind
- Measured: estimated from data
- Evidence
- Weak
- Plausible range
- 0.3 elasticity of compliance cost to income to 0.8 elasticity of compliance cost to income
Below 1, so compliance costs fall hardest, as a share of income, on small providers: the OfS's own fee schedule falls from £591 to £11 per student across its bands. 0.5 means a provider four times larger carries twice the cost.
Parameter regulation.scaleElasticity
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 2% of entrants at small and Level 4-5 providers to 10% of entrants at small and Level 4-5 providers
Stringent thresholds close this share of provision at smaller and Level 4-5 providers, where weak outcomes concentrate (the OfS has investigated franchised provision with poor outcomes, much of it at such providers); relaxed thresholds let it grow by as much.
No source: a judgement, open to challenge.
Parameter outcomes.entrantEffect
Share of compliance cost that is the TEF
10% of compliance cost
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 5% of compliance cost to 20% of compliance cost
TEF submissions are one exercise among many; its cost changes in proportion to the setting (abolished, simplified, as now, expanded).
No source: a judgement, open to challenge.
Parameter regulation.tefShareOfCompliance
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 1% of home undergraduate fee income to 4% of home undergraduate fee income
The part of providers' access and participation spending driven by regulation rather than their own aims; it changes by half for 'light' or 'intensive'. No sector-wide figure for providers' own access spending is published.
No source: a judgement, open to challenge.
Parameter access.costShareOfHomeFees
Politics (scenario)
How public mood moves with what the model produces, and when the government reacts — used only when politics is switched on. Scenario assumptions, not predictions.
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 100 mood points per proportional change to 400 mood points per proportional change
Concern is 50 at 2024-25's international numbers and reaches 100 at a fifth more. Most of the public supports international students, but their numbers count towards net migration, which governments have repeatedly acted to cut (the 2024 dependants rule, the levy, the shorter Graduate Route).
Parameter politics.migrationSensitivity
Concern about closures with none
20 mood points
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 10 mood points to 40 mood points
The starting level: the crisis has low salience until it becomes concrete.
No source: a judgement, open to challenge.
Parameter politics.closureBase
- Kind
- Assumption: a judgement
- Evidence
- Weak
- Plausible range
- 4 mood points per closure to 15 mood points per closure
Closures make the crisis concrete: seven in ten would worry if 30-40 universities closed, and most would blame the government. Concern reaches the reaction threshold after about seven closures.
Parameter politics.closurePerClosure
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 0.5 mood points per provider to 2 mood points per provider
Providers at serious risk above 2024-25's number, as reported in the press.
No source: a judgement, open to challenge.
Parameter politics.closurePerAtRisk
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 100 mood points per proportional change to 400 mood points per proportional change
Concern is 50 at the 2024-25 fee in real terms and reaches 100 at a fifth more.
No source: a judgement, open to challenge.
Parameter politics.feeSensitivity
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 100 mood points per proportional change to 400 mood points per proportional change
Concern is 50 at 2024-25's real public cost and reaches 100 at a fifth more.
No source: a judgement, open to challenge.
Parameter politics.spendingSensitivity
Mood at which the government may react
75 mood points
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 60 mood points to 90 mood points
Below this, nothing happens.
No source: a judgement, open to challenge.
Parameter politics.reactionThreshold
Chance a year that the government reacts when mood is at its height
40% chance a year
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 20% chance a year to 70% chance a year
The chance rises from nothing at the threshold to this at 100; each reaction happens at most once, the following academic year.
No source: a judgement, open to challenge.
Parameter politics.reactionChance
Cut to the teaching grant when concern about spending is high
20% of the grant
- Kind
- Assumption: a judgement
- Evidence
- None
- Plausible range
- 10% of the grant to 30% of the grant
The grant is the most discretionary large item in the ledger.
No source: a judgement, open to challenge.
Parameter politics.teachingGrantCut
Variation between providers
Year-to-year swings and the history each synthetic provider starts with, measured from real accounts.
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- 3.5% a year (standard deviation) to 7% a year (standard deviation)
Every provider's student numbers and other earned income get their own seeded random shocks each year, so providers diverge as real ones do. Measured on total income and applied to fees and other earned income (about 85 per cent of income), so the model's variation is slightly smaller than the measured one.
Parameter shocks.recruitmentSd
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- 0.1 autocorrelation of growth to 0.4 autocorrelation of growth
A good year tends to be followed by another; the shocks accumulate, so a provider's recruitment can drift well away from its group over ten years, as in the data.
Parameter shocks.recruitmentPersistence
Share of a provider's 2024-25 margin gap from its group that is lasting
86%
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- 75% to 95%
The 2024-25 accounts include one-off items. The model projects costs forward from 2024-25, so without this a one-off in that year would last for ever. From 2025-26, 14 per cent of each provider's margin gap from its group is removed (in its other operating costs); each group's totals are unchanged, because the gaps add up to zero.
- HE Provider Data: Finance, tables 1, 3, 4 and 12 (2015/16 to 2024/25), analysed in data/simulator/hesa-behaviour.json — HESA (Jisc). Provider margins relative to their group's median carry over at 0.686 after one year, 0.548 after two and 0.466 after three (1,870 and 1,618 pairs, 2016/17-2024/25, winsorised). A lasting part that closes about a fifth each year, plus a one-off part, fits all three: 0.86 x 0.80^k.
Parameter history.marginPersistentShare
- Kind
- Measured: estimated from data
- Evidence
- Moderate
- Plausible range
- 1.5% of income (standard deviation) to 3.5% of income (standard deviation)
Independent each year, seeded, and the same under every policy. Added to other operating costs. Together with recruitment shocks it is why providers move in and out of deficit as real ones do.
Parameter shocks.marginNoiseSd
Providers in deficit in 2024-25 that were also in deficit in 2023-24
65.2%
- Kind
- Measured: estimated from data
- Evidence
- Strong
- Plausible range
- 60% to 70%
Sets each synthetic provider's run of deficits at the start, so a provider already in its third deficit year acts sooner than one in its first.
Parameter history.deficitAlsoPriorYear
Providers in deficit in 2024-25, 2023-24 and 2022-23
48.3%
- Kind
- Measured: estimated from data
- Evidence
- Strong
- Plausible range
- 43% to 53%
As history.deficitAlsoPriorYear.
Parameter history.deficitAlsoTwoYearsBefore
Current policy
What the model treats as current policy: what is in force or announced, with the date each value takes effect.
Undergraduate tuition fee cap (approved fee providers)
9,250 from 2017-08-01; 9,535 from 2025-08-01; 9,790 from 2026-08-01; 10,050 from 2027-08-01; then indexed to RPI. (£ per student per year)
Indexation after 2027-28 is stated government intention, not yet law, and is promised only to 'high-quality' providers. The baseline assumes it happens for everyone; 'fees frozen from 2028' is one of the counterfactual presets. The fee cap applies to full-time undergraduates; foundation-year and part-time fees differ, so fee income per FTE is scaled by the cap rather than set equal to it.
International student levy
0 from 2017-08-01; 925 from 2028-08-01; then indexed to RPI. (£ per international student (headcount) per year)
Uprating is assumed to start in 2029-30, the year after introduction.
International students per provider exempt from the levy
220 from 2028-08-01. (students (headcount))
Stated intention; the figure will be fixed in secondary legislation.
OfS recurrent teaching grant (Strategic Priorities Grant), relative to 2024-25
1 from 2017-08-01; 0.955 from 2026-08-01. (index, 2024-25 = 1)
2024-25 to 2025-26 is assumed flat. No settlement exists beyond 2026-27; the baseline holds it in cash terms, so it falls in real terms. Applied to the part of each group's funding body grants that is not QR research funding.
Research England quality-related (QR) research funding, relative to 2024-25
1 from 2017-08-01; 1.019 from 2026-08-01; then indexed to CPI. (index, 2024-25 = 1)
After 2026-27 QR is assumed to grow with CPI, following the spending review statement that core QR grows 'in line with expected inflation'. Each group's QR is its share of £1,987m in proportion to its research grant income.
UKRI grants: share of full economic cost paid
0.8 from 2006-01-01. (fraction)
The nominal rate. TRAC shows institutions actually recovered 70.1 per cent of full economic cost on research council work in 2024-25. Used from Phase 3.
Employer National Insurance rate (Class 1 secondary)
0.138 from 2023-04-06; 0.15 from 2025-04-06. (fraction of earnings above the threshold)
The 13.8 per cent rate before April 2025 is the long-standing previous rate.
Employer National Insurance secondary threshold
9,100 from 2023-04-06; 5,000 from 2025-04-06. (£ per year)
The £9,100 threshold before April 2025 is the long-standing previous value. The employment allowance (£10,500) is ignored as immaterial for providers of this size.
Teachers' Pension Scheme employer contribution rate
0.237 from 2019-09-01; 0.287 from 2024-04-01; 0.177 from 2027-04-01. (fraction of pensionable pay)
Mostly affects post-1992 universities, whose academic staff are in the TPS. Unlike schools and colleges, universities received no government funding for the 2024 increase, and so keep the whole of the 2027 reduction.
Universities Superannuation Scheme employer contribution rate
0.145 from 2024-01-01. (fraction of salary)
Held at 14.5 per cent in the baseline. The outcome of the 2026 valuation is a candidate shock.
Graduate Route length (non-doctoral)
24 from 2021-07-01; 18 from 2027-01-01. (months)
Used from Phase 3, when international demand responds to policy.
Taught postgraduates may bring dependants
1 from 2010-01-01; 0 from 2024-01-01. (1 = yes, 0 = research degrees only)
Used from Phase 3.
Cap on international entrants, as a share of 2024-25 entrants
0 from 2017-08-01. (fraction (0 = no cap))
No cap exists. The setting exists so that the cap lever has a baseline to change.
Regulatory compliance burden, relative to 2024-25
1 from 2017-08-01. (index, 2024-25 = 1)
The 2024-25 burden is already inside every provider's costs; the lever changes it from 2027-28. Its cost per provider is set by the regulation.* parameters.
Policy on shared services and collaboration
0 from 2017-08-01. (0 = as now, 1 = facilitated, 2 = facilitated with VAT neutralised)
Shared services exist today on a small scale; their savings are already in the 2024-25 accounts. The lever adds wider adoption from 2027-28.
Presets and shocks
Presets are starting points: each sets some of the rules and leaves the rest at current policy. Historical presets apply past settings from 2026; they are not re-runs of history.
The Chapter 20 settlement
The book's own recommendations to government, as far as the model can express them: stop erosion of the fee and the teaching grants, remove the VAT penalty on collaboration, recover more of research's cost, regulate proportionately with earlier attention to weak finances, and build a restructuring regime before the collapse.
Sets: Teaching grant indexation Rises with inflation (CPI); Shared services and collaboration Facilitated, with the VAT penalty removed; UKRI share of full economic cost 90%; Regulatory burden Proportionate (a quarter less); Financial monitoring Preventative; Restructuring regime for failing providers Formal regime, set up in advance.
- The fee cap keeps the RPIX indexation already planned; the book asks for it to be made durable, which the model already assumes.
- International rules stay as they are: the book asks for stability and notice — published assessments, a full recruitment cycle before changes — not for particular settings.
- The book asks funders to state and improve the share of research cost they pay, without naming a rate; 90 per cent is an illustration.
- 'Proportionate' regulation is illustrated as a quarter less compliance cost; the book is explicit that this is not deregulation.
- Preventative financial monitoring stands for the chapter's call for heavier attention 'while intervention can still help'.
- The restructuring regime is modelled as earlier, convened searches for partners, orderly transfer of most students when a provider does close, public transfer funding, and a small premium charged by lenders — the cost government cited in rejecting it.
Fees frozen to 2036
The fee cap reaches £10,050 in 2027-28 and is then frozen, as it was from 2017 to 2024.
Sets: Fee cap indexation after 2027-28 Frozen.
Research pays its full cost
Research council grants pay the full economic cost of research, at the government's expense.
Sets: UKRI share of full economic cost 100%.
- The model keeps research volume fixed; with a fixed budget, the same rate would fund fewer projects.
International students capped
New international entrants capped at 20 per cent below 2024-25 levels from 2027-28.
Sets: Cap on international entrants 20% below 2024-25.
Low-regulation OfS
Compliance costs halved.
Sets: Regulatory burden Minimal.
- Any effect of lighter regulation on quality or student protection is not modelled.
Collaborative sector
Government pushes shared back-office services and neutralises the VAT that makes sharing dearer than duplication.
Sets: Shared services and collaboration Facilitated, with the VAT penalty removed.
International rules as in 2023
The 2023 rules applied from 2026: two-year Graduate Route, taught postgraduates' dependants allowed, no levy.
Sets: Graduate Route 24 months (until December 2026); Taught postgraduates' dependants Allowed, as before 2024; International student levy £0.
- Applied from 2026; not a re-run of what happened.
2012's fee in today's money
The £9,000 fee of 2012-13 restored to its real value: about £15,000 in 2027-28.
Sets: Undergraduate fee cap £15,000.
- The DfE estimates the 2025-26 fee of £9,535 is worth about £5,920 in 2012-13 prices, so £9,000 then is about £14,500 in 2025-26 money and over £15,000 by 2027-28; £15,000 is the top of the lever.
- The model assumes higher fees do not deter home applicants; the government's loan cost rises with the fee.
The parties' proposals
Five presets set the rules to what the parties contesting elections across England have proposed for universities: the Labour government's current policy, and the published proposals of the Conservatives, the Greens, the Liberal Democrats and Reform UK (in alphabetical order). The model covers England; fees and university regulation are devolved, so parties that stand only in Scotland or Wales are not included.
Each preset moves a control only where a party has said something the model can represent, and says how. Where a party has stated an aim but not how it would work — which courses lose places, what replaces fee income, how large a grant — the preset makes a choice and flags it as the model's assumption, not the party's. Proposals the model cannot represent, mostly changes to what graduates repay, are listed with the reason. Every change starts in 2027-28, as for every control, although a government elected at the next general election (due by August 2029) would act later. Positions are as at the date given; parties' policies change, and corrections are welcome.
Labour (the government)
Current policy: the Labour government's settled and announced plans. Its White Paper's other aims for universities come without figures, so the model cannot represent them beyond current policy.
- Proposed: The fee cap rises with inflation: £10,050 in 2027-28, then with forecast RPIX. (Department for Education, 2025-11-21). Not modelled: Already current policy, so already in every run.
- Proposed: A levy of £925 a year for each international student from August 2028, with an allowance for each provider's first 220. (Department for Education, 2026-07-13). Not modelled: Already current policy, so already in every run.
- Proposed: Targeted maintenance grants for disadvantaged students on courses that support the government's missions and Industrial Strategy. (Department for Education, 2025-11-21). Not modelled: Already current policy. The model counts only changes to maintenance grants, so their cost is not in its public cost.
- Proposed: Improve the cost recovery of research. (Department for Education, 2025-11-21). Not modelled: No rate is given. The research cost-recovery control can test one.
- Proposed: Encourage greater specialisation, based on research and teaching strengths, in collaboration with other institutions. (Department for Education, 2025-11-21). Not modelled: No mechanism or scale is given. The shared services and merger controls can test some.
Checked 2026-10-05.
Conservative (opposition)
About 100,000 fewer university entrants a year, through number controls aimed at courses with poor graduate outcomes, with the savings spent on apprenticeships.
Sets: Cut in home undergraduate places 22%; Which courses lose places Mostly courses with low graduate earnings; Emphasis on graduate earnings Strong: courses judged on earnings.
- Proposed: Number controls on every subject group, cutting about 100,000 university places a year. (Times Higher Education, 2025-10-08). In the model: 100,000 is about 22 per cent of a year's home undergraduate entrants in England.
- Proposed: The controls fall on courses that consistently lead to poor graduate outcomes. (Times Higher Education, 2025-10-08). The model's assumption: Poor outcomes are read as low graduate earnings (LEO), both in where places are cut and in how courses are judged. The party has not said how courses would be chosen.
- Proposed: Cap the interest on Plan 2 student loans at RPI, removing the extra of up to 3 per cent. (PA Media (via AOL), 2026-02-22). Not modelled: Plan 2 covers students who started between 2012 and 2023. Their repayments do not change universities' income, and the model counts only the cost of new lending.
- Proposed: Spend the savings (put at £3.6bn) on doubling apprenticeships. (PA Media (via AOL), 2026-02-22). Not modelled: Apprenticeships are outside the model. The saving on student loans appears in its public cost.
Checked 2026-10-05.
Green (opposition)
Tuition fees abolished and maintenance grants restored. The party has not said how universities would be funded instead; the preset assumes the government pays them the fee in full.
Sets: Tuition fees abolished Yes: government pays providers instead; Maintenance support £3,500; Taught postgraduates' dependants Allowed, as before 2024.
- Proposed: Scrap tuition fees. (Green Party, 2024-07-02). The model's assumption: The party has not said what replaces fee income. The preset has the government pay each provider the fee cap (£10,050 in 2027-28, rising as planned) for every home undergraduate. Set the fee cap lower to test a smaller grant.
- Proposed: Restore maintenance grants for undergraduates. (Green Party, 2024-07-02). The model's assumption: No amount is given. The preset uses £3,500 for each eligible student, about the last maximum grant (£3,387 in 2015-16), as for the Liberal Democrats.
- Proposed: Allow international students to bring dependants. (Higher Education Policy Institute, 2024-06-10). In the model: Dependants allowed for taught postgraduates again, as before 2024.
- Proposed: In the long run, cancel graduate debt. (Higher Education Policy Institute, 2024-06-10). Not modelled: Changes what graduates repay, not what universities receive.
- Proposed: Spend £30bn more on research by the end of the Parliament. (Higher Education Policy Institute, 2024-06-10). Not modelled: The split between universities and other research, and the timing, are not given.
Checked 2026-10-05.
Liberal Democrat (opposition)
Maintenance grants restored, the international student levy and the employer National Insurance rise reversed, shared services encouraged and research budgets raised with inflation.
Sets: Maintenance support £3,500; International student levy £0; Employer National Insurance 13.8%; Shared services and collaboration Actively facilitated; Research council budget 105%.
- Proposed: Reintroduce maintenance grants on top of student loans, aiming to cover 35 per cent of undergraduates, rising to 50 per cent when public finances allow. (Liberal Democrats, 2026-03). The model's assumption: No amount is given. The preset uses £3,500 for each eligible student, about the last maximum grant (£3,387 in 2015-16).
- Proposed: Reverse Labour's National Insurance rise and the international student levy. (Liberal Democrats, 2026-03). In the model: The levy is never charged, and the employer rate returns to 13.8 per cent. The lower threshold that came with the rise is not modelled.
- Proposed: Explore shared-service reform. (Liberal Democrats, 2026-03). The model's assumption: 'Explore' is read as actively facilitating shared services, without removing the VAT penalty.
- Proposed: Increase research budgets with inflation and move towards multi-year settlements. (Liberal Democrats, 2026-03). The model's assumption: The model's current plans have research grants falling behind inflation by about half a per cent a year, because much of UKRI's budget is flat in cash. Keeping pace with inflation would leave them about 5 per cent higher by 2035-36. The preset uses 5 per cent from 2027-28, slightly generous in the early years.
- Proposed: Interest-free loans for internal restructuring. (Liberal Democrats, 2026-03). Not modelled: The model has no government lending to providers. The restructuring regime is a different policy.
- Proposed: Unfreeze graduate repayment and parental income thresholds; write off some debt for key public-service workers after 10 years. (Liberal Democrats, 2026-03). Not modelled: These change what graduates repay, not what universities receive.
- Proposed: Double degree apprenticeships in skill-shortage areas. (Liberal Democrats, 2026-03). Not modelled: Apprenticeships are outside the model.
Checked 2026-10-05.
Shocks are stress tests applied in a year of your choice. A policy facing shocks is compared with current policy facing the same shocks; with no rules changed, the shocks are compared with none. Their sizes are stress-test choices, not forecasts.
International recruitment falls by 30%
A sudden, lasting loss of international demand — a visa scandal, a currency crisis in a main market, a competitor's offer.
Evidence for its size: weak. Dependant applications fell by almost 80 per cent in the first quarter after the 2024 restriction; a 30 per cent fall in all international entrants is a severe but not unprecedented stress for individual markets.
Inflation of 8%
An inflation spike like 2022-23's: prices jump, pay catches up only partly, and the fee cap — uprated on a forecast set in advance — follows a year late.
Evidence for its size: weak. Modelled on 2022-23, scaled down: national pay awards then were 3 per cent plus one-off payments, well below inflation, before catching up. Because the model uprates the fee cap with the RPI of the same year, the shock moves the RPI spike a year later to reflect fees set on forecasts made in advance.
A recession
Two years in which earned income (commercial, residences, conferences) and international demand fall, while more home students apply, as they did in 2008-09.
Evidence for its size: weak. Home demand for higher education tends to rise when jobs are scarce; the sizes are illustrative.
A five-year fee freeze
The fee cap is held in cash terms for five years from the chosen year, then indexation resumes from the frozen level.
Evidence for its size: strong. The mechanism is exact; whether a government would do it is the scenario. The cap was frozen at £9,250 from 2017-18 to 2024-25.
Research funding cut by 15%
Research council grants, other research income and QR all fall by 15 per cent for good.
Evidence for its size: weak. An illustrative cut in the public research budget, passed on to universities' research income. Providers do not cut their research activity automatically: they respond only as their finances require.
Home applications fall by 10%
A lasting fall in home demand — a weaker graduate premium, a shift to apprenticeships, or cost-of-living pressure.
Evidence for its size: weak. Illustrative. Applies to new entrants, so it reaches students on roll over the length of a course.
Interest rates up 3 points
Borrowing costs rise by three percentage points on the share of the sector's debt that floats or is refinanced.
Evidence for its size: none. Most university debt is fixed-rate, which the model assumes elsewhere; 30 per cent floating or due for refinancing is an assumption. Higher returns on cash, which would partly offset the cost, are not modelled.
A major insolvency
The weakest large or medium-sized university closes, and the shock spreads: international applicants grow warier and lenders and merger partners more cautious for three years.
Evidence for its size: none. Untested in England, so every part of this is an assumption: which provider fails (the one with least cash in its group), and how far confidence falls.
How uncertain the results are
Every result the simulator shows first is a single run on the model's central assumptions. "How sure is this?" runs 200 more. Each draws every parameter that has a plausible range from a triangular distribution over it — most likely at the central value, never outside the range — and draws fresh year-to-year swings in recruitment and one-off items. Assumptions you have set yourself stay fixed. The synthetic sector stays the same, so each building's share of runs in distress can be shown.
When you compare a policy with current policy, both are run on the same draws, so the range of the differenceis narrower than the range of either: much of the uncertainty affects both alike. "Better in 80 per cent of runs" means the policy came out ahead on that measure in four draws out of five.
These are shares of the model's runs, not probabilities of what will happen. Most ranges are judgements; each assumption is drawn independently, although in reality inflation, pay, recruitment and costs move together; the runs share one synthetic sector; and the ranges cannot cover a relationship the model leaves out, or ranges that are themselves wrong. The ranges show how much the results depend on the model's assumptions, within the model.