A graduate employment rate is the result of a series of choices: who counts as a graduate, when they are surveyed, what counts as a job, how further study and part-time work are treated, who answers and what happens to those who do not. Change any of these and the same cohort produces a different percentage. Raw rates from different universities or countries are therefore not comparable unless those choices are identical and published. Before comparing two rates, readers should check six things: population, timing, definition, denominator, coverage and change over time.
Six decisions behind every employment rate
- Population. All graduates, or only some levels, only domestic graduates, or only those living in the country?
- Timing. Four months, nine months, fifteen months or several years after completion?
- Definition of work. Any paid work in a reference week, or full-time work only? Does self-employment count?
- Denominator. All graduates, all respondents, or only those available for work? Are graduates in further study removed, counted as a positive outcome, or reported separately?
- Coverage. What share of graduates is represented, and do the people who answered differ from those who did not?
- Stability. Have the questions, definitions or data sources changed between years?
The rest of this article shows, with published examples, how each decision moves the number.
What counts as "employed"?
The international standard is deliberately broad. Under the resolution of the International Conference of Labour Statisticians (as amended in 2023), persons in employment include those who worked in a job for pay or profit for at least one hour during a short reference period. The resolution explains that the one-hour criterion is meant to cover part-time, temporary, casual or sporadic work. That is appropriate for labor force statistics, but it means that an "employment rate" built on this logic tells readers nothing about whether graduates have enough work.
Australia's Graduate Outcomes Survey, part of the government's Quality Indicators for Learning and Teaching (QILT), shows how definitions can be layered. QILT defines overall employment as one or more hours per week and full-time employment as 35 hours or more per week across all jobs. Its comparison measures express the full-time rate as a percentage of graduates available for full-time employment, and the overall employment rate as a percentage of graduates available for employment. A university quoting the overall rate and another quoting the full-time rate may be describing very similar graduates with very different numbers.
The European indicator uses yet another frame. Eurostat's employment rate of recent graduates covers people aged 20 to 34 who completed their highest level of education one to three years before the survey interview and who are no longer in formal or non-formal education or training. On that basis, Eurostat reports a rate of 87.0 percent for recent tertiary graduates in the EU in 2025 (data extracted June 2026). Graduates still studying are outside the denominator by design.
When is "after graduation"?
National instruments measure at very different points. Lessky and Dollinger (2025, Journal of Higher Education Policy and Management) compared three graduate outcome instruments in detail. They report that Australia's Graduate Outcomes Survey measures short-term outcomes about four to six months after completion, that the UK's Graduate Outcomes survey collects data about 15 months after completion, and that the European Eurograduate survey covers graduates about 12 to 15 months and five years after completion. In Ireland, the Central Statistics Office (CSO) describes the national Graduate Outcomes Survey as reaching graduates about nine months after graduation.
| Measure | Timing after completion | Notable scope rule | Source |
|---|---|---|---|
| Australia, Graduate Outcomes Survey | About 4 to 6 months | Published national results refer to domestic graduates; full-time rate uses graduates available for full-time work | QILT; Lessky and Dollinger (2025) |
| Australia, longitudinal follow-up | About 3 years | Re-surveys graduates who completed the earlier survey; domestic results published | QILT |
| Ireland, Graduate Outcomes Survey | About 9 months | All graduates of levels 6 to 10, including international graduates | CSO, hosted by the Higher Education Authority |
| UK, Graduate Outcomes | About 15 months | Course and placement details linked from the student record rather than asked | Lessky and Dollinger (2025) |
| Europe, Eurograduate | About 12 to 15 months and 5 years | Core module common to all 17 participating countries; other modules optional | Lessky and Dollinger (2025) |
| EU, recent graduates indicator | 1 to 3 years | Ages 20 to 34; people still in education or training excluded | Eurostat |
Timing alone can move a rate a long way. QILT reports that the full-time employment rate of domestic undergraduates immediately after graduation was 80.2 percent for the 2023 cohort and 91.8 percent when the same graduates were surveyed about three years later in 2026, a rise of 11.6 percentage points. Comparing a four-month figure from one place with a fifteen-month figure from another compares two stages of a career, not two systems.
Lessky and Dollinger also show that instruments differ in what they ask, not only when. Their comparison found significant differences in how graduate outcomes are measured across the three instruments. As the authors put it, "There is no such thing as a neutral or objective survey design." Each instrument embeds assumptions about what matters.
Who answers, and who is missing?
Most graduate outcome figures come from surveys, and most surveys miss many graduates. Senekal and Munro (2019, South African Journal of Higher Education) reviewed 23 graduate tracer studies from 13 countries published between 1995 and 2016. Of the 21 studies for which response rates were relevant, 12 had response rates below 50 percent. The authors point out that a response rate is not the same as representativeness: what matters is whether those who respond differ from those who do not. A meta-analysis by Groves and Peytcheva (2008, Public Opinion Quarterly) pooled 59 methodological studies designed to estimate nonresponse bias, including studies that matched administrative records to sample members, and, according to its abstract, examined under which circumstances nonresponse rates are related to nonresponse bias and under which they are not. A response rate is therefore a warning light, not a measure of bias.
The Irish CSO offers an unusual direct check. It linked Graduate Outcomes Survey records to its Educational Longitudinal Database, which draws on higher education, social protection and tax records, for about 317,000 graduates (about 278,000 when tax and social protection data were added). It reports an average response rate of 47 percent across five survey years (2018 to 2023, excluding 2020), with about 150,000 responses. Response rates were above 50 percent among graduates whom the administrative data placed in substantial employment, were lower in NFQ levels 6 and 7 than in levels 8 to 10, and were lowest among graduates working in accommodation and food services. The CSO concluded that response rates were good overall, with no unusual results. Its analysis also illustrates a quieter form of missing data: comparing survey answers with tax records, the CSO suggests that graduates on lower salaries may have been more likely to answer "I'd rather not say" to the salary question. If that is the case, our inference is that survey-based salary figures would lean upward even when the response rate is good.
Survey versus records: one institution's comparison
A peer-reviewed case study in The AIR Professional File (Bryant, 2021) compared two methods for the same cohorts at Western Washington University. The earlier method combined a graduate survey with searches of a professional networking site for graduates with recent employment records. The case study notes that this supplementary search could only find positive outcomes, so it selected part of the sample on the outcome being measured. The newer method used state wage records and enrollment records for graduates living in the state.
| Measure | 2014-15, survey method | 2014-15, wage records | 2015-16, survey method | 2015-16, wage records |
|---|---|---|---|---|
| Graduates in cohort | 3,174 | 3,174 | 3,308 | 3,308 |
| Share of cohort with data | 42.7% | 93.4% | 45.2% | 92.3% |
| Employed | 82.7% | 75.2% | 83.1% | 75.6% |
| Further education | 11.9% | 12.0% | 13.2% | 12.9% |
| Salaries reported | 296 | 1,945 | 100 | 2,011 |
In both years the survey-based method produced an employment rate 7.5 percentage points higher than the records-based method. The two methods also define employment differently. In the records method, a graduate counts as employed if they have positive wage records in the second quarter after graduation and are not enrolled in further education. Because state wage records exclude federal employees and the self-employed, the case study adjusts the employment rate upward using statewide ratios for those groups. The comparison therefore does not isolate nonresponse bias. It shows that method choices together can move a published rate by several points, and that a salary figure based on a few hundred voluntary responses may describe a small, self-selected group.
Further study, part-time work and self-employment
These three categories are where definitions diverge most.
- Further study. Eurostat excludes people still in education or training from its recent-graduate indicator. The records method in Bryant's case study classifies graduates enrolled within three quarters of graduation as seeking further education and does not analyze their wages. The Irish administrative classification has separate categories for substantial employment only and for substantial employment combined with education. A rate that quietly counts further study as a positive outcome is not comparable with one that removes it.
- Part-time work. Broad employment definitions include part-time and casual work. Australia's separate full-time measure exists for this reason. Rates that do not state whether part-time work is included cannot be read with confidence.
- Self-employment. Administrative data may or may not see it. The Irish administrative definition of substantial employment includes self-employment, whereas the US state wage records in Bryant's case study do not, which is why the case study needed an adjustment. Tang and colleagues' analysis of Australian graduate survey data excluded self-employed occupations from its analytic samples. Each choice is defensible; mixing them is not.
Administrative definitions also differ from survey ones. The CSO notes that in its database "substantial employment" means at least 12 weeks of insurable work in a calendar year, whereas the survey records what a graduate is doing at the moment of response, so the two sets of categories are not directly comparable even for the same people.
International graduates: often invisible in the data
International graduates are frequently the group with the weakest coverage. Tang, Perales, Rowe and Baxter (2022, Journal of Population Research) analyzed the Australian Graduate Survey for 2000, 2008 and 2015, measuring outcomes about four months after course completion. Response rates were 41 to 43 percent for international graduates and 58 to 63 percent for local graduates, and the authors note that international graduates who remained in Australia are most likely over-represented among respondents. In 2015, among international graduates in the analytic sample, 32.7 percent were unemployed (not working but actively seeking work), compared with 10.0 percent of domestic graduates; the international figure had been 9.8 percent in 2000. The authors attribute much of the deterioration to a falling share of international graduates who were Australian citizens or permanent residents. They also note that the national survey was replaced in 2016 by an instrument that is not directly comparable, which limits trend analysis.
Administrative data do not solve the problem automatically. The CSO notes that internationally domiciled graduates who never obtained an Irish personal public service number are not covered by tax and social protection records, and that graduates with no activity in any administrative source are classed as "not captured," most of whom are assumed to have emigrated or returned to their country of origin. In Australia, QILT's published national results refer to domestic graduates only. A university-wide rate may therefore describe mainly domestic graduates even when a large share of its graduates are international.
When the definition changes, the trend changes
Even a single national series can shift for reasons unrelated to graduates. QILT reports that the undergraduate full-time employment rate rose from 74.0 percent in 2024 to 75.4 percent in 2025, and states that these movements are likely attributable to a change in how the labor force was defined in 2025. The longitudinal follow-up page also records that an earlier version of the 2026 national report contained errors in selected figures and commentary and was corrected the next day. Both are examples of good practice: the publisher told readers. A year-on-year change of one or two points should never be read as a real change in graduate outcomes until definitional changes and corrections have been ruled out.
A comparability checklist for graduate employment rates
The checklist below is an editorial framework proposed by this article. It is not the UNIRANKS methodology and does not describe how UNIRANKS evaluates institutions. It is intended for anyone reading or publishing a graduate employment rate.
| Check | Question to ask | Why it matters |
|---|---|---|
| Population | Which graduates are included, and which are excluded (levels, domestic or international, residents only)? | Excluding hard-to-reach groups can raise or lower a rate |
| Reference date | How many months after completion, and in which year? | Rates rise substantially over the first years after graduation |
| Employment definition | Any paid work, a minimum number of hours, or full-time only? Is self-employment included? | The broadest definition counts one hour of paid work |
| Denominator | All graduates, all respondents, or only those available for work? How is further study treated? | Removing graduates in study or not seeking work changes the base |
| Coverage | What share of the cohort is represented, and have respondents been compared with nonrespondents or with records? | Responding graduates may differ systematically from others |
| Missing items | How many respondents skipped salary or job questions? | Item nonresponse can bias averages even with a good response rate |
| Uncertainty | Are counts and confidence intervals shown, especially for small programs? | Small groups produce volatile percentages |
| Stability | Have questions, definitions or data sources changed since the last release? | A method change can look like a trend |
If two rates do not match on the first four checks, they should not be compared at all. If they match on those but differ on coverage, any comparison should be stated with explicit caution.
Our interpretation
The examples above come from different countries and institutions, but they point in the same direction. Each decision behind a rate can move it by several percentage points or more: 7.5 percentage points between two methods for the same cohorts in one case study, 11.6 points between two measurement points for the same Australian cohort, and a national rise that the publisher itself attributes largely to a definitional change. None of these sources shows that one method is correct for all purposes. Linked administrative records improve coverage but miss some forms of work and people who leave the country; surveys capture job content and graduates' own views but depend on who answers. Our reading is that a rate without its definition, timing, denominator and coverage should be treated as descriptive information about the graduates who were counted, not as a measure of institutional quality.
Practical steps
For anyone comparing published rates
- Look for the reference date and definition before looking at the percentage.
- Prefer full-time or hours-based measures over "any employment" when judging whether graduates have enough work.
- Check whether international graduates are included and how well they are covered.
- Compare rates only within the same instrument, year and definition, and by field of study where possible.
For universities publishing rates
- Publish the population, reference date, definition, denominator, response rate and number of respondents next to every rate.
- Report full-time employment, part-time employment, further study and other activities separately instead of one combined "positive outcome" figure.
- Where lawful and available, compare survey respondents with administrative or enrollment records, as the Irish CSO did, and publish what the comparison shows.
- Never supplement survey data with sources that can only find successful graduates.
- Flag every change in method or definition in the same place as the figure it affects.
Measuring employment status is only one part of measuring employability. For the wider picture, including job quality, field-of-study alignment and career progression, see How Should Universities Measure Graduate Employability?
Limitations
This article draws on five peer-reviewed sources. Four were read in full (Tang and colleagues; Lessky and Dollinger; Senekal and Munro; Bryant) and one from its abstract only (Groves and Peytcheva). Groves and Peytcheva (2008) and Senekal and Munro (2019) are older than the other sources and are used for methodological points that do not depend on recent data. The examples come mainly from Australia, Ireland, the United States and Europe; the institutional comparison describes a single university whose graduates mostly stay in one state, and the author notes that the method may not work where graduates are more dispersed. The Australian international-graduate figures date from 2015 and describe a survey that has since been replaced. The official statistics cited describe their own systems and are not presented as universal standards. The checklist is editorial and has not been validated.
Related reading on UNIRANKS
More research on graduate outcomes is collected in the Employability & Career Success research hub. Support for international students during study and the transition to work is discussed in What Support Should Universities Provide International Students?
Editorial disclosure
UNIRANKS publishes university rankings, and its methodology includes an Employability & Career Success dimension. This article reviews independent research on graduate employment rates and their comparability. It does not describe how UNIRANKS scores institutions.
Research References
- Bryant, M. (2021). Using State Workforce Data to Report Graduate Outcomes. The AIR Professional File, Article 154. https://doi.org/10.34315/apf1542021 (verified 22 November 2026)
- Groves, R. M., & Peytcheva, E. (2008). The Impact of Nonresponse Rates on Nonresponse Bias: A Meta-Analysis. Public Opinion Quarterly, 72(2), 167-189. https://doi.org/10.1093/poq/nfn011 (verified 22 November 2026)
- Lessky, F., & Dollinger, M. (2025). An international comparison of graduate outcome survey instruments: a critical reflection. Journal of Higher Education Policy and Management, 47(6), 707-728. https://doi.org/10.1080/1360080X.2025.2491884 (verified 22 November 2026)
- Senekal, J., & Munro, N. (2019). Lessons learnt from two decades of graduate tracer research: Recommendations for the South African context. South African Journal of Higher Education, 33(2), 230-248. https://doi.org/10.20853/33-2-2628 (verified 22 November 2026)
- Tang, A., Perales, F., Rowe, F., et al. (2022). From bad to worse: examining the deteriorating labour market outcomes of international graduates in Australia. Journal of Population Research, 39(3), 441-473. https://doi.org/10.1007/s12546-022-09291-7 (verified 22 November 2026)
Sources
- Graduate Outcome Survey - Data Validity Assessment (June 2024). Central Statistics Office, published by the Higher Education Authority (Ireland). https://hea.ie/assets/uploads/2025/07/Graduate-Outcomes-Survey-Data-Validity-Assessment-CSO-PDF.pdf (accessed 22 November 2026)
- Employment rates of recent graduates (Statistics Explained). Eurostat. https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Employment_rates_of_recent_graduates (accessed 22 November 2026)
- Graduate Outcomes Survey. Quality Indicators for Learning and Teaching (QILT), Australian Government. https://qilt.edu.au/surveys/graduate-outcomes-survey-(gos) (accessed 22 November 2026)
- Graduate Outcomes Survey - Longitudinal. Quality Indicators for Learning and Teaching (QILT), Australian Government. https://qilt.edu.au/surveys/graduate-outcomes-survey---longitudinal-(gos-l) (accessed 22 November 2026)
- Resolution to amend the 19th ICLS resolution concerning statistics of work, employment and labour underutilization (Resolution II, 21st International Conference of Labour Statisticians, 2023). International Labour Organization. https://www.ilo.org/sites/default/files/wcmsp5/groups/public/@dgreports/@stat/documents/normativeinstrument/wcms_230304.pdf (accessed 22 November 2026)
