UniRank UK

UK Graduate Outcomes: How LEO and Graduate Outcomes Data Feed the Ranking Tables

LEO and the HESA Graduate Outcomes survey are the two datasets behind graduate-outcomes columns in UK league tables. Here is how each is built, how old the numbers really are, and how to check the sample size before you trust a provider's figure.

A graduate-outcomes figure in a league table can look more precise than the underlying evidence warrants. Its meaning depends on when graduates were observed, which destinations count, how small groups are handled and whether the figure describes an individual course or a pooled subject area.

What LEO actually measures

The Longitudinal Education Outcomes dataset, usually abbreviated to LEO, links education history with later employment, earnings and benefits records. Education history includes the schools, colleges and higher education providers attended, the subjects studied and the qualifications obtained. Discover Uni describes LEO as using government tax records.

LEO is therefore an administrative dataset built by linking records. It is not built by surveying graduates about their salaries or employment.

The observation date is crucial. In the Department for Education’s 2023-24 LEO release, graduates from the 2017-18 cohort have employment and earnings outcomes recorded during the 2023-24 tax year, five years after graduation. The release also examines people who completed GCSEs in 2012-13 against their outcomes in the same tax year, ten years after their GCSEs. A publication date or tax year is therefore not the cohort’s graduation year.

Discover Uni states that its displayed LEO data covers graduates who finished their studies between 2015 and 2017. Even when a current table refers to “LEO”, the underlying group may have left university several years earlier. Its figure describes recorded outcomes for that cohort, not the salary available to a graduate today.

LEO is also subject-based rather than course-based. The 2023-24 release presents outcomes for 35 subject areas at CAH2 level and provides a more detailed CAH3 breakdown. It does not establish how every individual course within one of those subject areas performed.

The figures are revised. Each annual LEO dataset adds another year of information but may also revise earlier years because newer administrative data has become available. A figure quoted in an older article may consequently differ from the most recently revised version without the underlying methodology having changed in the same way.

Coverage also needs checking. The wider LEO study applies to England, Scotland and Wales, while the cited 2023-24 release covers first-degree and postgraduate graduates from higher education providers in England. Discover Uni separately states that Northern Ireland is excluded because LEO does not contain data for Northern Irish graduates.

How a graduate prospects score is calculated

LEO should not be confused with HESA’s Graduate Outcomes survey. They use different sources and observation periods.

In The Complete University Guide’s published methodology, the “graduate prospects – outcomes” measure comes from the Graduate Outcomes survey conducted fifteen months after graduation for students in the HESA 2022-23 data.

Its calculation is straightforward:

  • The numerator is graduates entering employment or further study.
  • Employment included in that measure is highly skilled employment, rather than lower-level jobs that do not normally recruit graduates.
  • The denominator is graduates with a known destination.

That denominator is the critical qualification. Graduates without a known destination are outside the calculation. The published percentage therefore describes the proportion with a known destination who entered employment or further study, not automatically the proportion of every graduate who did so. A precise percentage can conceal a substantial difference between these two groups.

This survey-based graduate prospects score is also not a five-year earnings measure. LEO follows a historical cohort through administrative records at a later point; graduate prospects – outcomes records destinations soon after graduation. Their numbers should not be read as substitutes for one another.

A separate measure called “graduate prospects – on track” uses another survey response. It is the proportion who agreed or strongly agreed with the statement “My current activity fits with my future plans”. It should not be treated as either an earnings figure or the highly skilled employment measure.

Before interpreting any table, check its stated input. The documented graduate prospects methodology uses HESA’s Graduate Outcomes survey, while LEO supplies longer-term earnings information at subject-area level on Discover Uni. A league table’s use of graduate-outcomes information does not establish that every earnings or destination measure in it comes from LEO.

Apply the sample-size test

The Office for Students says Discover Uni shows how many students contribute to published data and, for survey data, the response rate. Those fields need to be read alongside the result.

Its publication thresholds are explicit:

  • For most data, it needs at least 10 students.
  • For LEO, it needs at least 15 students.
  • Survey data may also be withheld when response rates are low.

The thresholds protect individual identities and help guard against publishing results based on groups that are too small to interpret reliably. A larger group combined with a higher response rate gives greater confidence than a small or incompletely answered group. Where numbers are smaller, small differences between providers or subjects should not be treated as decisive.

LEO is not published for individual courses. For smaller courses, data may be grouped with other courses in the same subject area at that university or college. This is particularly important when comparing two subject-level figures. The displayed number may pool several courses, so it cannot reveal whether the result came from one course, a combination of courses or a change in the mix of students between them.

Before accepting a figure, find:

  • the number of students contributing to it;
  • the response rate if it comes from a survey;
  • the cohort years;
  • the subject-area grouping; and
  • whether the figure is suppressed, adjusted or weighted.

If those details are absent, the number is difficult to compare responsibly.

Interpretation limits built into the data

Annualised LEO earnings do not fully account for whether graduates work full-time or part-time. The data combines both, so earnings can appear lower where more graduates work part-time. The Department for Education supplies separate illustrative full-time-equivalent salary estimates based on Labour Force Survey data. It also reports that the gender earnings gap is significantly smaller once hours worked are accounted for.

Discover Uni presents median salaries, which describe the middle of a group rather than an individual prediction. A median can still combine graduates working different hours and in different employment patterns.

League-table aggregation can add another layer. In The Complete University Guide methodology, two years of continuation and graduate prospects data are combined. Scores are withheld where the number of graduates is too small, and weighting is reduced in tables with a small range of data so that graduate prospects do not exert a disproportionate effect on the overall ranking. These are table-specific rules, not universal rules for every league table.

A blank cell in a subject table is also not a zero score. It indicates that the required data or measure is unavailable. Treating it as zero would incorrectly convert missing information into a negative result.

Checklist before comparing figures

For any graduate-outcomes table in front of you, check:

  • Source: Is it LEO, HESA’s Graduate Outcomes survey or another dataset?
  • Time: Which cohort is observed, and how long after leaving university?
  • Population: Which qualifications, regions and providers are covered?
  • Base: How many students contribute, and what is the survey response rate?
  • Grouping: Is the figure for a course, a subject area or a pooled group?
  • Definition: Are unknown destinations outside the denominator?
  • Adjustment: Are earnings annualised without a full adjustment for hours, and is the measure suppressed, aggregated or downweighted?

These datasets describe group outcomes rather than individual futures. In particular, LEO does not show the causal effect of studying at a particular qualification level, and a median historical outcome is not a forecast of what one graduate will earn.