How the numbers work

Salary data

Graduate starting salaries and the 5-year and 10-year benchmarks are drawn from the DfE Longitudinal Educational Outcomes (LEO) dataset, published annually by the Department for Education. LEO is uniquely reliable because it links HMRC tax records directly to graduate records — it reports what graduates actually earned, not what they told a survey they earned.

We use median earnings (50th percentile) at approximately 1 year, 5 years, and 10 years after graduation, broken down by HECoS subject group. For professions with nationally set pay scales — NHS doctors (Agenda for Change / Medical and Dental pay review) and nurses (Agenda for Change bands 5–7) — we use the published NHS England pay scales for 2023/24 as the starting anchor, cross-checked against LEO.

The ONS Annual Survey of Hours and Earnings (ASHE) 2025, Table 20 (Annual pay — Gross, age group by 2-digit SOC occupation) is used for two purposes: alt-path (non-graduate) earnings benchmarks, and as a cross-check on senior earnings trajectories beyond 10 YAG. Each field cites its SOC code inline in the source data. ASHE is a 1% sample of PAYE records — employer-reported, not self-reported — making it directly comparable in quality to LEO.

Important caveat for humanities and social science fields:LEO reports earnings for everyone who graduated with a given degree, regardless of what job they subsequently entered. For subjects with diverse graduate destinations — History, Sociology, Languages, Psychology, Politics — the median reflects a population scattered across law, finance, teaching, civil service, and many other sectors. The number is accurate for the typical graduate; it is not a prediction of earnings specifically in a "history career." We flag this on each relevant degree page.

Student loan repayment model

The calculator models the UK student loan system as it applies to English-domiciled students. We support Plan 2 (started university 2012–2023) and Plan 5 (starting 2023 onwards). Scottish, Welsh, and Northern Irish students have different plans — we flag this in the UI and the model does not yet cover them.

RPI is modelled at 3.0% p.a. as a conservative long-run assumption (Bank of England target is 2% CPI; RPI historically runs 0.5–1 pp above CPI). Salary growth within each career field uses field-specific compound rates implied by the LEO progression data. We do not model income tax or National Insurance — repayments are calculated on gross earnings above the threshold as per SLC rules.

AI-automation risk

AI risk scores (1 = highly resilient, 10 = high exposure) reflect occupational automation exposure at the task level, not the job level. A score of 8 for Marketing doesn't mean marketing jobs will disappear — it means the specific tasks that dominate entry-level marketing roles (first-draft copy, basic analysis, reporting) are the tasks that current AI tools already perform competently.

Our scores are derived from three inputs, weighted qualitatively:

1. Oxford Martin Programme— Frey & Osborne's original occupational automation probability scores, updated via their 2023 reassessment accounting for large language model capabilities.
2. OECD Employment Outlook 2023— "AI and the Labour Market" chapter, which applies a task-level automation framework to OECD occupational data including UK SOC codes.
3. ONS "Which occupations are at highest risk of automation?" (2019, with 2023 update) — UK-specific task analysis using O*NET activity data mapped to SOC 2010.

Confidence ratings (High / Medium / Low) reflect how stable we expect the score to be over the next 3–5 years, not certainty in the current estimate. High = regulatory or physical constraints make the score unlikely to shift significantly; Low = the technology is moving fast enough that the entry-level task mix for this field could look materially different within a degree cycle.

Limitations

• Medians hide distributions. Law and finance are highly bimodal — the median salary is not representative of either the City track or the high-street track. The calculator shows the median; your individual outcome depends heavily on institution, specialisation, and employer.
• No parental means-testing. Maintenance loan entitlement varies by household income. We model the maximum maintenance loan available for a student living away from home outside London.
• Threshold uplift not guaranteed. The repayment threshold is indexed to average earnings, but Parliament can change this. Plan 2 thresholds were frozen from 2021–2025. We model thresholds as fixed in real terms; if they are frozen in nominal terms (as happened), actual repayments will be higher than modelled.
• No career interruptions. The model assumes continuous employment above the threshold from graduation. Periods of part-time work, career breaks, or unemployment lower actual repayments and extend the loan lifecycle.
• England only. Plan 2 applies to English-domiciled students who started between 2012–2023. Students from Scotland, Wales, and Northern Ireland, and those who started before 2012 (Plan 1) or after 2023 (Plan 5), should use the relevant plan.

Data refresh schedule

LEO data is published each autumn (typically September/October). We update salary figures annually following each LEO release. AI risk scores are reviewed annually against new OECD and ONS automation publications. The current data vintage is 2023/24 tax year — sourced directly from cah3_subject_level_data.csv in the DfE LEO 2023-24 release, using 1, 5, and 10 years after graduation (YAG) median earnings, weighted across CAH3 subject groups within each CAH2 category.

If you spot an error or have a source that contradicts a figure here, please email data@degreeindex.co.uk.