Comparing the performance of on-campus and online students in an accidental experiment

This guest post comes from Patrick Bijsmans (Maastricht University) & Arjan H. Schakel (University of Bergen)

Students and staff are experiencing challenging times, but, as Winston Churchill famously said, “never let a good crisis go to waste”. Patrick recently led a new undergraduate course on academic research at Maastricht University (read more about the course here). Due to COVID-19 students could choose whether they preferred online or on-campus teaching, which resulted in 10 online groups and 11 on-campus groups. We were presented with an opportunity to compare the performance of students who took the very same course, but did so either on-campus or online. Our key lesson: particularly focus on online students and their learning.

In exploring  this topic, we build on our previous research on the importance of attendance in problem-based learning, which suggests that students’ attendance may have an effect on students’ achievements independent from students’ characteristics (i.e. teaching and teachers matter, something that has also been suggested by other scholars). We created an anonymised dataset consisting of students’ attendance, the number of intermediate small research and writing tasks that they had handed in, students’ membership of an on-campus or online group, and, of course, their final course grade. The latter consisted of a short research proposal graded Fail, Pass or Excellent.

316 international students took the course, of which 169 (53%) took the course online and 147 (47%) on-campus. 255 submitted a research proposal, of which 75% passed. One of the reasons why students did so well – normal passing rates are about 65% – might be that, given that this was a new course, the example final exam that they were given was one written by the course coordinator. Bolkan and Goodboy suggest that students tend to copy examples, so providing them may therefore not necessarily be a good thing. Yet students had also done well in previous courses, with the cohort seemingly being very motivated to do well despite the circumstances.

But on closer look it’s very telling that 31% of the online students (52 out of 169) did not receive a grade, i.e. they did not submit a research proposal. This was 9.5% for the on-campus students (14 out of 147)[1]. Perhaps this is the result of self-selection, with motivated students having opted for on-campus teaching. Anyhow, it is clear that online teaching impacts on study progress and enhancing participation in examination among online students needs to be prioritised by programme directors and course leaders.

We focus on students that at least attended one meeting (maximum 6) and handed-in at least one assignment (maximum of 7). Out of these 239 students, 109 were online students (46%) and 130 on-campus (54%). Interestingly, on average these 239 students behaved quite similarly across the online and on-campus groups, they attended on average 5 meetings (online: 4.9; on-campus: 5.3) and they handed-in an average of 5 to 6 tasks (online: 5.0; on-campus: 5.9).

We ran a logit model with a simply dummy variable as the dependent variable which taps whether a student passed for the course. As independent variables we included the total number of attended meetings and the total number of tasks that were handed-in. Both variables were interacted with a dummy variable that tracked whether students follow online or offline teaching and we clustered standard errors by 21 tutor groups.

Unfortunately, we could not include control variables such age, gender, nationality and country of pre-education. This would have helped to rule out alternative explanations and to get more insight into what factors drive differences in performance between online and offline students. For example, international students may have been more likely to opt for online teaching and may have been confronted with time-zone differences, language issues, or other problems.

Figure 1: Impact of attending class on the probability to pass

Figure 1 displays the impact of attending class on the probability to pass for the final research proposal. The predicted probabilities are calculated for an average student that handed-in 5 tasks. Our first main finding is that attendance did not matter for online students, but it did for on-campus students. The differences in predicted probabilities for attending 3, 4, 5, or 6 meetings are not statistically significant (at the 95% confidence level) for online students but they are for on-campus students. Students who attended the maximum of six on-campus meetings had a 68% higher probability to pass compared to a student who attended 3 meetings (89% versus 21%) and a 52% higher probability to pass compared to a student who attended 4 meetings (89% versus 37%).

Figure 2. Impact of handing-in tasks on the probability to pass

Figure 2 displays the impact of handing-in tasks on the probability to pass for the final research proposal. The predicted probabilities are calculated for an average student that attended 5 online or on-campus meetings. Our second main finding is that handing-in tasks did not matter for on-campus students, but it did for online students. The differences in predicted probabilities for handing-in 4, 5, 6, or 7 tasks are not statistically significant (at the 95% confidence level) for on-campus students but they are for online students. Students who handed-in the maximum of seven tasks had a 51% higher probability to pass compared to a student who handed in four tasks (69% versus 18%) and a 16% higher probability to pass compared to a student who handed-in five tasks (69% versus 53%).

Note that we do not think that attendance does not matter for online students or that handing-in tasks does not matter for offline students. Our dataset does not include a sufficient number of students to expose these impacts. From our previous research we know that in general we can isolate the impact of various aspects of course design with data from three cohorts (around 900 students). The very fact that we find remarkably clear-cut impacts of attendance among on-campus students and of handing-in tasks for online students for a relatively small number of students (less than 240) reveals that these impacts are so strong that they surface and become statistically significant in such a small dataset as ours.

This is why we feel confident to advise programme directors and course leaders to focus on online students. As Alexandra Mihai also recently wrote, it is worth investing time and energy in enhancing online students participation in final examinations and to offer them many different small assignments to be handed-in during the whole time span of the course. This is not to say that no attention should be given to on-campus students and their participation in meetings but, given limited resources and the amount of gain to be achieved among online students, we think it would be wise to first focus on online students.


[1] The difference of 21% in no grades between online and offline students is statistically significant at the 99%-level (t = 4.78, p < 0.000, N = 314 students).