Attendance shows that a student was present. It does not show that the student was thinking, participating, connected to others or learning. Student engagement is better measured as a set of separate layers: presence (attendance), recorded activity (logins, submissions, contributions), engagement itself (behavioral, cognitive and emotional investment in learning) and learning outcomes (what students can demonstrate). Research supports using several sources together: attendance records, learning-platform data, validated engagement surveys and assessment evidence, read separately and disaggregated for part-time, commuting and online students, whose engagement often happens away from the timetable and the campus.
Four measures that are often treated as one
Institutional reports frequently use "engagement" as a label for whatever data are easiest to collect. The first step in measuring engagement well is to keep four ideas apart.
- Attendance records physical or virtual presence at a scheduled session. It is a necessary condition for some kinds of learning, but it says nothing about what happened during the session.
- Activity records actions captured by systems: learning management system (LMS) logins, page views, video plays, forum posts, library loans and submissions. It is abundant and cheap, but an action is not the same as effort or understanding.
- Engagement is a psychological and behavioral construct. Most research divides it into behavioral engagement (participation, effort, persistence), cognitive engagement (deep processing, self-regulation, connecting ideas) and affective or emotional engagement (interest, belonging, enthusiasm or, at the other end, frustration).
- Learning is what students know and can do as a result, demonstrated through valid assessment. Engagement is usually studied as a route to learning, not as a substitute for it.
Each of these can rise while another falls. A student can attend every lecture while disengaged, or attend rarely while studying intensively from recordings and readings. Treating one layer as evidence for another is the most common measurement error in this area.
| Measure | What it records | What it cannot show | Groups it can misrepresent |
|---|---|---|---|
| Attendance registers or swipe data | Presence at scheduled sessions | Attention, participation, independent study, learning | Part-time, commuting and online students, students with caring or work commitments |
| LMS and digital activity logs | Clicks, logins, views, posts, submissions | Quality of thinking, offline study, emotional engagement | Students who download materials and study offline, or who use other tools |
| Engagement surveys | Self-reported study behaviors, interactions and perceptions | Verified behavior; views of students who do not respond | Groups with low response rates |
| Assessment results | Demonstrated learning against criteria | The process that produced it; prior preparation | Students who entered with different prior attainment |
What the research shows
Attendance predicts grades, but how students attend matters
The most cited quantitative synthesis on attendance is a meta-analysis by Credé, Roch and Kieszczynka (2010, Review of Educational Research). According to its abstract, attendance was strongly related to class grades (69 samples, 21,195 students, meta-analytic correlation 0.44) and to grade point average (33 samples, 9,243 students, 0.41), and it was relatively independent of admissions test scores and high school grades. Mandatory attendance policies were associated with a small positive effect on average grades, but that estimate rested on only three studies with 1,421 students. These are correlations: they show that attendance is a useful early signal, not that requiring attendance produces learning.
Büchele (2021, Assessment & Evaluation in Higher Education) asked whether a simple attendance count is enough. The abstract notes that studies generally find positive but mostly weak correlations between attendance and performance, and that the study examined classroom engagement dimensions as mediating factors. Its conclusion is that "it does not matter if but rather how students attend class." Only the abstract was available for this article, so the sample and effect sizes are not reported here.
A before-and-after study by Edwards and Clinton (2019, Higher Education) shows how quickly attendance can change when the learning environment changes. In a compulsory second-year undergraduate research methods module within a three-year UK degree, matched cohorts before (161 students) and after (160 students) the introduction of lecture capture were compared across three matched lectures. The share of students who attended none of the three rose from 19.9% to 39.4%. Attendance predicted attainment after controlling for previous grades and gender, while viewing of recordings showed no significant relationship with attainment once attendance was taken into account. In the year with recordings, 68% of the cohort did not view lecture capture to any significant degree during the term. The authors acknowledge that the two cohorts may have differed in unobserved ways and that the result may not generalize across subjects and institutions. For measurement, the lesson is that a fall in attendance after a change in delivery is not automatically a fall in engagement, and recording use cannot simply be counted as an equivalent substitute.
Activity data are plentiful but weak and course-specific
Conijn, Snijders, Kleingeld and Matzat (2017, IEEE Transactions on Learning Technologies) analyzed LMS data from 17 blended undergraduate courses with 4,989 students at a single Dutch university of technology. Prediction models built from LMS variables varied strongly across courses even within one institution, so models did not transfer well from one course to another. Once interim assessment grades were available, LMS data added little predictive value, and they were also of limited use for early intervention. The authors note that the available courses were mostly first-year courses with similar blended designs. Click counts, in other words, are a noisy indicator whose meaning depends on how each course uses the platform.
A systematic evidence map by Bond, Buntins, Bedenlier, Zawacki-Richter and Kerres (2020, International Journal of Educational Technology in Higher Education) reviewed 243 studies on educational technology and student engagement in higher education published between 2007 and 2016. Only 18 studies (7%) defined student engagement; 225 (93%) did not. Behavioral engagement was documented in 209 studies (86%), affective engagement in 163 (67%) and cognitive engagement in 136 (56%). Among the most frequent "engagement indicators" in the corpus was achievement itself, which illustrates how often engagement and learning are merged. The authors also note that surveys can capture unobservable aspects such as feelings and cognitive strategies but can be lengthy and subject to self-selection, and they suggest combining observation with learning analytics data. The review covered English-language studies only, mostly from the United States and the United Kingdom.
Engagement is multidimensional, and its definition is contested
Wong and Liem (2022, Educational Psychology Review) describe the construct as conceptually hazy and group the problems into four areas: overgeneralization, "jingle-jangle" fallacies (the same label used for different things, or different labels for the same thing), ambiguity about what the student is engaged with, and under-theorization. They propose distinguishing engagement in learning from engagement with the institution. Their review is framed around schooling and educational psychology rather than higher education specifically, and only the abstract was read here.
Kahu and Nelson (2018, Higher Education Research & Development) place engagement within what they call the educational interface, the space where student and institutional factors meet. In their refined framework, four psychosocial constructs (self-efficacy, emotions, belonging and well-being) are proposed as mechanisms linking student and institutional characteristics to engagement and success. The framework is conceptual rather than an empirical test, but it explains why two students with the same demographic profile and the same attendance record can have very different experiences, and it argues for evaluating interventions against these mechanisms.
How engagement surveys approach the problem
Large engagement surveys try to measure the behaviors and conditions that research links to learning, rather than satisfaction. The National Survey of Student Engagement (NSSE), run by a research center at Indiana University, reports ten Engagement Indicators grouped into four themes: Academic Challenge (higher-order learning, reflective and integrative learning, learning strategies, quantitative reasoning), Learning with Peers (collaborative learning, discussions with diverse others), Experiences with Faculty (student-faculty interaction, effective teaching practices) and Campus Environment (quality of interactions, supportive environment). Each indicator summarizes several related questions and is expressed on a 60-point scale. NSSE's own guidance warns that "individual student scores vary much more within institutions than do average scores between institutions," so an institutional average hides most of the variation that matters for support.
Australia's Student Experience Survey (SES), whose indicators are published on the Quality Indicators for Learning and Teaching (QILT) website, includes a peer engagement indicator built from five questions about how often students contributed to class discussion, worked with other students, gave or received study help, interacted socially and interacted with a diverse range of students. Its examples explicitly include online chat forums and discussion boards, which matters for students who rarely attend in person.
Surveys capture cognitive and emotional engagement that logs cannot, but they depend on self-report and on who responds. How to read survey percentages, response rates and wording is covered in the companion article on what student satisfaction survey results really mean.
Students who are not on campus every day
Commuting students
Turner, Webb and Pritchard (2025, Higher Education) used survey data from more than 500 first-year students in the UK to compare those who relocated for university with those who commuted. According to the abstract, commuter students appeared more engaged on the measures used, showing positive study behaviors and readiness to work with course mates, and online learning supported continuity when they were away from campus. Time on campus had little association with engagement among commuters, whereas less time on campus was associated with lower engagement among relocated students. The authors interpret the findings through the concept of mattering. Only the abstract was read, and a first-year survey cannot show how engagement develops later. Even so, it is a direct warning against equating time on campus with engagement.
Online students
Martin and Borup (2022, Educational Psychologist) note that reported learner engagement is lower online than in person and argue for reconceptualizing online engagement around both its dimensions (cognitive, affective, behavioral) and the affordances of the online environment. They identify five research themes: engagement through communication, interaction, presence, collaboration and community. For measurement, this suggests looking at the quality of interaction and presence in online courses, not only at the volume of logins. Only the abstract was read.
Part-time students
No study reviewed for this article isolated part-time students, so the following is interpretation rather than evidence. Part-time students typically spend fewer hours on campus and in scheduled sessions by design. Any indicator built on hours present, events attended or co-curricular participation will therefore tend to show them as less engaged even when their engagement with course content is high. The frameworks above suggest a fairer comparison: measure engagement per unit of study load, ask about cognitive and peer engagement within their courses, and report part-time results separately rather than inside an institutional average.
A layered framework for evidence
The table below is an editorial framework proposed by this article; it is not the UNIRANKS methodology and does not describe how UNIRANKS evaluates institutions. It sets out what each layer of evidence can support and how to avoid penalizing students whose study pattern differs from the full-time, on-campus norm.
| Layer | Example evidence | What it can support | Main validity risk | Adjustment for part-time, commuting and online students |
|---|---|---|---|---|
| 1. Presence | Attendance at scheduled sessions, in person or live online | Early warning of disconnection when attendance drops sharply | Confuses presence with engagement; affected by recordings and timetable clashes | Compare against each student's own expected schedule and study load, not a full-time norm |
| 2. Activity | LMS logins, submissions, forum posts, library use | Patterns of participation within a single course design | Course-specific meaning; offline study invisible; models transfer poorly | Interpret within the course; do not compare raw counts across modes of study |
| 3a. Behavioral engagement | Survey items on effort, participation, collaboration; observed contribution | Whether students take part in practices associated with learning | Self-report and nonresponse bias | Include online collaboration and asynchronous discussion as participation |
| 3b. Cognitive engagement | Survey items on higher-order learning, reflection, learning strategies; quality of written work | Depth of study | Hard to observe; relies on validated instruments | Usually independent of location, so a fair basis for comparison |
| 3c. Affective engagement | Belonging, interest, mattering, frustration | Conditions that may sustain or undermine engagement | Mood and context effects; small subgroups | Ask whether students feel they matter, not only whether they attend events |
| 4. Learning | Assessment results, progression, value-added measures | What students can demonstrate | Prior attainment and assessment design confound comparisons | Report by mode of study with prior attainment taken into account |
Our interpretation
Three conclusions follow from reading these studies together. First, attendance remains a useful signal, and the meta-analytic association with grades is substantial, but the more recent work suggests that the quality of participation matters more than the count. An attendance target is therefore a weak engagement goal on its own. Second, digital activity data are attractive because they are complete and continuous, yet the evidence suggests they are noisy, course-dependent and of limited additional value once assessment information exists. They are better used to prompt a conversation with a student than to describe an institution. Third, the students most likely to be misread are those who study differently: commuters, online learners and part-time students. The commuter study is a single first-year survey, but it reverses the common assumption that less time on campus means less engagement.
For anyone comparing institutions, the practical consequence is that an "engagement rate" without a stated definition, data source and population is uninterpretable. Engagement can also be shaped by learning spaces and facilities, a topic covered in the UNIRANKS article on campus facilities and student success.
Practical steps
For university administrators and analysts
- Define the layer before collecting the data. State whether an indicator measures presence, activity, engagement or learning, and label reports accordingly.
- Use attendance and activity as triggers, not verdicts. A sudden drop is a reason to contact a student, not evidence of disengagement in itself.
- Validate activity metrics course by course. Check whether a platform indicator relates to outcomes in the specific course before using it, as the variation reported by Conijn and colleagues suggests.
- Add a validated engagement instrument. Use survey items on cognitive and peer engagement, and report response rates and nonrespondent characteristics.
- Disaggregate by mode of study. Report part-time, commuting and online students separately, and judge them against their own expected patterns.
- Keep learning evidence separate. Do not count grades as engagement; link the two only in analyses that account for prior attainment.
For students
When comparing universities, ask how engagement is supported for students in your situation: whether sessions are recorded and how recordings are meant to be used, whether online discussion and group work are part of courses, and whether part-time or commuting students have timetables and support that fit their circumstances.
For researchers
Define engagement explicitly, specify which dimension is measured, avoid treating achievement as an engagement indicator, and report results for non-traditional study patterns. The evidence map by Bond and colleagues found that most studies in its corpus did none of the first three.
Limitations
This article draws on nine peer-reviewed sources. Three were read in full (Bond and colleagues; Conijn and colleagues; Edwards and Clinton) and six from their abstracts only (Büchele; Credé, Roch and Kieszczynka; Kahu and Nelson; Martin and Borup; Turner, Webb and Pritchard; Wong and Liem). Where only the abstract was read, no sample sizes, methods or statistics beyond the abstract are reported. The meta-analysis on attendance dates from 2010 and predates widespread lecture recording; the lecture capture and LMS studies each come from a single institution; and the commuter study is a first-year survey from one country. No study reviewed here focused on part-time students, so the section on them is interpretation. Almost all the evidence is correlational. The framework is editorial and has not been validated as a measurement instrument.
Related reading on UNIRANKS
For the wider context, see the Student Experience & Satisfaction research hub and the companion article on how universities should measure student experience. Measured teaching and learning outcomes are discussed in the Quality of Education section.
Editorial disclosure
UNIRANKS publishes university rankings, and its methodology includes a Student Experience & Satisfaction dimension. This article reviews independent research on measuring student engagement. It does not describe how UNIRANKS scores institutions.
Research References
- Bond, M., Buntins, K., Bedenlier, S., et al. (2020). Mapping research in student engagement and educational technology in higher education: a systematic evidence map. International Journal of Educational Technology in Higher Education, 17(1), 2. https://doi.org/10.1186/s41239-019-0176-8 (verified 22 November 2026)
- Büchele, S. (2021). Evaluating the link between attendance and performance in higher education: the role of classroom engagement dimensions. Assessment & Evaluation in Higher Education, 46(1), 132-150. https://doi.org/10.1080/02602938.2020.1754330 (verified 22 November 2026)
- Conijn, R., Snijders, C., Kleingeld, A., et al. (2017). Predicting Student Performance from LMS Data: A Comparison of 17 Blended Courses Using Moodle LMS. IEEE Transactions on Learning Technologies, 10(1), 17-29. https://doi.org/10.1109/TLT.2016.2616312 (verified 22 November 2026)
- Credé, M., Roch, S. G., & Kieszczynka, U. M. (2010). Class Attendance in College. Review of Educational Research, 80(2), 272-295. https://doi.org/10.3102/0034654310362998 (verified 22 November 2026)
- Edwards, M. R., & Clinton, M. E. (2019). A study exploring the impact of lecture capture availability and lecture capture usage on student attendance and attainment. Higher Education, 77(3), 403-421. https://doi.org/10.1007/s10734-018-0275-9 (verified 22 November 2026)
- Kahu, E. R., & Nelson, K. (2018). Student engagement in the educational interface: understanding the mechanisms of student success. Higher Education Research & Development, 37(1), 58-71. https://doi.org/10.1080/07294360.2017.1344197 (verified 22 November 2026)
- Martin, F., & Borup, J. (2022). Online learner engagement: Conceptual definitions, research themes, and supportive practices. Educational Psychologist, 57(3), 162-177. https://doi.org/10.1080/00461520.2022.2089147 (verified 22 November 2026)
- Turner, R., Webb, O., & Pritchard, C. (2025). Academic engagement amongst commuter students: a comparative analysis. Higher Education, 90(3), 761-775. https://doi.org/10.1007/s10734-024-01349-0 (verified 22 November 2026)
- Wong, Z. Y., & Liem, G. A. D. (2022). Student Engagement: Current State of the Construct, Conceptual Refinement, and Future Research Directions. Educational Psychology Review, 34(1), 107-138. https://doi.org/10.1007/s10648-021-09628-3 (verified 22 November 2026)
Sources
- Engagement Indicators. National Survey of Student Engagement, Indiana University Center for Postsecondary Research. https://nsse.indiana.edu/nsse/survey-instruments/engagement-indicators.html (accessed 22 November 2026)
- Student Experience Survey (SES). Quality Indicators for Learning and Teaching (QILT). https://www.qilt.edu.au/surveys/student-experience-survey-(ses) (accessed 22 November 2026)
