1 Introduction

An established literature shows that students in higher education who are taught in-person outperform those who are taught online, and the difference is large [7, 10, 20, 21, 35]. For example, Bettinger et al. [6] found that taking a course online reduced students’ course grades by almost 0.33 of a standard deviation. While the magnitude should be treated with caution, due to selection, this is supported by a growing experimental literature—where students are randomly assigned to online or in-person lectures—that shows that students benefit from in-person teaching [1, 4, 14, 19] although there are some exceptions [2]. Despite the benefits from in-person teaching almost one third of courses in UK universities are taught via online lectures—up from 4 percent before the pandemic [36]. While students are the ones who bear the cost of online teaching (via lower attainment) the use of online lectures is often justified by the argument that this is what students prefer.

Before the Covid-19 pandemic, the literature generally found that students preferred in-person teaching [18, 37] although there is considerable variation across institutions and programmes [12, 15]. Since the start of the Covid-19 pandemic there is growing evidence that students’ preferences have shifted slightly toward online teaching [27, 32, 38] although this literature is not conclusive [5, 34]. However, much of this literature investigates preferences across specific programmes (e.g., Atwa et al. [5], Mudenda et al. [27] focus on medical students’) or Covid-19 specific cohorts (e.g., Melo et al. [26], Simonova et al. [34], Zhang et al. [38]) and it not obvious that the results from these studies generalise more broadly to university students today. We contribute to this literature by providing new evidence on students’ preferences—from two UK universities—over lecture mode across two distinct dimensions (expected learning and enjoyment).

In this paper, we use new experimental data from two UK universities to investigate students’ preferences regarding lecture mode and length, and assessment type. Our main findings are as follows. First, students hold strong preferences against online lectures. They expect to learn more from, and enjoy, modules that have in-person lectures. Second, students enjoy shorter lectures (1 h) more than longer lectures (2 h). Interestingly, students do not expect to learn more from longer lectures—suggesting diminishing marginal returns. Third, students prefer individual assignments to exams. They expect to enjoy and learn more from modules that have individual assignments (vs exams). All three of these results are robust across samples and all groups we consider.

Our paper also contributes to a growing body of work seeking to determine students’ preferred other aspects of module design such as lecture length and assessment method (e.g. Meginnis and Campbell [24]). Prior work on students’ preferred lecture length is mixed. While the literature consistently shows that students’ hold higher expected returns to lectures than other university related activities (e.g., Andrietti and Velasco [3], Delavande et al. [11]) the findings on students’ preferred lecture length are inconclusive [8, 16, 22]. We contribute to this literature by providing new evidence on students’ preferences for lecture length for both in-person and online lectures.

Our paper also contributes to a growing literature investigating students’ preferences for assessment types. The literature generally finds that students dislike exams [25, 30], and this is consistent across disciplines including economics [33] and mathematics [31]. We contribute to this literature by providing new evidence on students’ preferences across different assessment types in the context of their expected enjoyment and expected learning. This distinction is important because, if students dislike exams but expect to learn a lot from them, it would be easy to justify their inclusion in a module from a student perspective.

In this paper, we identify students’ preferences using experimental data, which represents an important contribution. The existing empirical literature generally identifies students’ preferences using a combination of the choices that they make (i.e., what modules they enrol onto) and modules satisfaction scores [13, 28, 29]. However, it can be challenging identifying students’ preferences using this data. First, different elements of module design are often highly correlated (i.e., 1-h lectures and individual assignments) so it is difficult to isolate the role of each individual element. Second, the different elements of module design are not randomly assigned—it is possible that other correlated factors are driving students’ preferences (i.e., teacher quality or module topic). By using new experimental data, we are able to overcome these challenges and identify students’ preferences over lecture mode, lecture length, and assessment type.

Understanding current students’ preferences also has important practical implications. Universities, and educators more generally, are increasingly assessed based on students’ subjective evaluations (e.g., the National Student Survey (NSS)). Therefore, understanding students’ preferences regarding different elements of module design could be advantageous—for example, boosting satisfaction scores by avoiding the methods they enjoy least.

This paper is organised as follows. Section 2 introduces our sample; Sect. 3 introduces our methods; Sect. 4 investigates students’ preferences while Sect. 5 shows that we can replicate these results with a second sample of students, Sect. 6 concludes.

2 Sample

The data we use comes from a survey of students at a leading UK business school, fielded in early November 2023.Footnote 1 The online survey took about 20 min to complete, and participation was incentivised with a £5 Amazon voucher. The survey was designed to collect information on students’ personal characteristics (e.g., sex, age, nationality, and ethnicity), family background (e.g., parents education level), and academic background (e.g., level of study, expected marks). In addition, students took part in an experiment designed to investigate preferences over different aspects of a module.

Students were recruited using two approaches. In two programmes (Masters in Management and Masters in Research), participation in the survey was mandatory and took place in class—all of students on these programs participated. For the other Programmes students were recruited by administrative staff circulating the survey information via the school’s intranet. We restricted participation to students who were studying full-time. A total of 463 students took part in the survey (approximately 15% of eligible students). We drop 104 students who did not participate in the experiment we use in this paper, leaving us with a sample of 359.Footnote 2 Sample characterises are shown in Table 1. As Table 1 shows, most of our sample are postgraduate (87%) and international (89%) students, which is what we’d expect given that participation was mandatory in two highly international postgraduate programmes.

Table 1 Shows the distribution of our two samples of students across observable characteristics (percentage) and differences by degree programme (Sample 1) and level of study (Sample 2)
Full size table

3 Methodology

Our survey participants took part in an experiment designed to isolate students’ preferences over module design. The factors that we considered were lecture mode, lecture length, and assessment method. Each of these factors could take on several values.Footnote 3

Each participant was presented with twelve hypothetical scenarios where they are asked to report their expected enjoyment and expected learning on a 0 to 100 scale. In the first eight scenarios, students were shown one individual factor. In the following four, students were shown all factors, with values randomly assigned. An example of these scenarios is presented in the online supplementary material. In total, each of our 359 participants are asked to report their expected enjoyment and expected learning twelve times, providing us with a total sample of 4207 and 4195 observations for learning and enjoyment respectively. The response rates were 98 percent and 97 percent.

Using the responses to these hypothetical scenarios, we investigate the effect that these factors have on enjoyment/learning (\({Y}_{is}\)) of student \(i\) in response to scenario \(s\), which contains factor(s) \(f\), using the following OLS specification:

$${Y}_{isf}= {B}_{0}+ {B}_{1}{\partial }_{f}+{\delta }_{i}+{\varepsilon }_{i}$$
(1)

Here, \({B}_{1}\) is our coefficient of interest, which demotes the percentage point effect of our factor fixed effects \({\partial }_{f}\). We also include student fixed effects \({\delta }_{i}\). We regress students’ enjoyment and expected learning on the factors fixed effects separately. Our standard errors are clustered at the individual level.

4 Results

Our results are presented in Table 2 (columns 1–4) and Fig. 1. Our three main findings are as follows: First, looking at lecture mode, students expect to enjoy, and learn more from, in-person lectures and the difference is large. Students report 10.5 percentage points (0.43sd) lower expected leaning and 6.6 percentage points (0.25sd) lower expected enjoyment for an online lecture compared to an in-person lecture of the same duration (one hour).

Table 2 Shows the coefficients for our factor fixed effects on students expected learning (Columns 1–2 and 5–6) and expected enjoyment (columns 3–4 and 7–8) measured on a 0–100 scale for students in Sample 1 (columns 1–4) and Sample 2 (columns 5–8)
Full size table
Fig. 1
Fig. 1
Full size image

Shows the coefficients for our factor fixed effects on students expected learning (blue) and expected enjoyment (red) measured on a 0–100 scale for students in Sample 1. For the Lecture mode and length, the omitted category is a one hour in-person lecture. For the assessment method the omitted category is an exam

Second, looking at lecture length, we find that students enjoy one-hour lectures more than two-hour lectures, with the difference being particularly pronounced for online teaching. Interestingly, students expect to learn the same amount (in-person) or less (online) in two-hour lectures compared to one-hour lectures of the same mode. Recall that students are asked about their preferences for factors individually and as part of a hypothetical module. In the latter total teaching time is held constant—any increase in lecture duration corresponds to an equivalent decrease in seminar time (in-person). There is no difference in students’ expected learning for online lectures by lecture length in response to the hypothetical modules compared to the individual factors (14 percentage points lower). Our interpretation of this result is that students’ lower expected learning for longer online lectures is driven by expected negative marginal returns and not by replacing an hour of in-person seminars with an additional hour of online lecture (see Delavande et al. [11] for a review of how students beliefs and expectations about the returns to different academic and non-academic activities influence behaviour at University in our setting).

Third, looking at assessment methods, we find that students have strong preferences against exams. They are the joint least enjoyable (with group projects) and have the joint lowest expected learning (with participation). In contrast, students have strong preferences for individual assignments. Individual assignments are the joint most enjoyable (with participation) and have the highest expected learning by a significant margin (8.6 percentage points or 0.35sd higher than exams).

While these results provide evidence on students on preferences (expected learning and enjoyment), we are also interested if this is consistent with the choices students make. Although it is challenging to use actual module choices to causally identify the effects of different elements of module design due to endogeneity, we can use a second experiment from the survey to investigate this directly. Participants also took part in a discrete choice experiment (or choice- based conjoint analysis) designed to identify the effect that different factors (lecture mode and assessment type) have on students’ module selection (see Appendix A2 for more details). The results from this experiment are consistent with our main findings. These results are presented in the supplementary material (Appendix Table 1 and Appendix Table 2). We find that students are more likely to select a module with an in-person lecture (vs an online lecture) and a module with an individual assignment (vs an exam). Moreover, the effect sizes are very large (1.2sd and 1.85sd respectively) and consistent across all the groups we consider.

Finally, we investigate heterogeneity in preferences by gender and degree programme (see Figs. 1 and 2 in the online supplementary material). The only differences according to these categories are that students on different degree programme (excluding MiM) prefer exams to group projects and dislike two-hour online lectures less than their MiM counterparts. While it is plausible that students in different programmes hold varying preferences, these differences could also be due to selection—the entire cohort of MiM students are represented. We will investigate the generalisability of these findings in the following section.

Fig. 2
Fig. 2
Full size image

Shows the coefficients for our factor fixed effects on students expected learning (blue) and expected enjoyment (red) measured on a 0–100 scale for students in Sample 2. For the Lecture mode and length, the omitted category is a one hour in-person lecture. For the assessment method the omitted category is an exam

5 Replication

Recall that our sample is from a highly selective UK business school. While our sample does include an entire cohort of students from one programme (Masters in Management), and these students preferences are generally similar to the other students in our sample, we cannot be certain whether these results generalise across levels of study (our students are predominately postgraduates) or institutions.

Conducting the same survey on a second sample of students from a research-intensive UK university, recruited using the EssexLab online subject pool in early December 2023, allows us to investigate this directly. A total of 512 participants took part in the survey. We drop 9 for not meeting the eligibility criteria (non-students) and 126 students who did not participate in the experiment we use in this paper, resulting in a final sample of 377. Sample characterises are shown in Table 1.

Our results using Sample 2 are presented in Table 2 (columns 5–8) and Fig. 2. These results are largely consistent, but there are two main differences. First, the magnitude of the preferences with respect to lecture mode differs—students in the second sample also dislike online lectures, but the magnitude is smaller (7.7 percentage points vs 10.2 percentage points). Second, students in the second sample prefer exams to group projects, whereas students in the first sample are indifferent between these two assessment types.

One benefit of online lectures is their compatibility with busy schedules. It is plausible that students who have other commitments might expect to enjoy, or learn more, from online lectures than from in-person lectures. As almost half of our sample have other commitments (i.e., paid work) we can investigate this directly (Fig. 3 in the online supplementary material). There is no difference in expected learning or enjoyment by students’ employment status.

Finally, we investigate heterogeneity in preferences by gender, nationality, and stage (see Fig. 4 and Fig. 5 in the online supplementary material). We observe heterogeneous preferences according to these categories for both lecture mode and assessment method. Postgraduate (international) students dislike online lectures more than undergraduates (home) and expect to learn less from them. This finding is consistent with the difference in magnitude observed between the two samples—the first sample is predominately postgraduate (international). Possibly reflecting gender differences in how work can be assigned in teams, we observe that female students enjoy group work significantly less than male students. There are no gender differences in students’ expected learning.

6 Discussion and concluding remarks

In this paper, we use new experimental data to investigate students’ preferences over lecture mode, length, and assessment type. Our main findings are as follows. First, students have strong preferences against online lectures. They expect to enjoy and learn more from modules with in-person lectures compared to modules with online lectures. Historically, students have preferred in-person teaching, but there was a shift in preferences during and immediately after the Covid-19 pandemic. These results suggest that students still favour in-person teaching and indicate that the increasing use of online lectures (4 percent of UK courses pre-covid to 30 percent today) is not necessarily well justified from a student satisfaction perspective. One reason for the shift to online teaching is cost reduction. In England, tuition fees have only increased by £250 per year since 2012, rising from £9000 to £9250 in 2017, meaning that home students are currently taught at a loss [17]. As our research suggests that students prefer in-person teaching, this finding could be used to justify raising tuition fees. Future work investigating students’ willingness to pay for in-person teaching could have significant policy implications.

Second, students enjoy shorter lectures (1 h) more than longer lectures (2 h). Interestingly, students do not expect to learn more from longer lectures, suggesting decreasing marginal returns. Third, students prefer individual assignments to exams. They expect to enjoy and learn more from modules that have individual assignments to exams. All three of these results are robust across two distinct samples of UK university students—a top ranked business school and a mid-ranked research-intensive university—and all groups within these samples, including an entire cohort of master’s students. These findings are also consistent with our findings from a discrete choice experiment.

One interpretation of these results is to guide module planning and design. When an educator is indifferent between assessment methods or lecture delivery modes due to pedagogical reasons and experience, it would make sense for them to use the approach that students are likely to prefer. Alternatively, these results could suggest using students’ preferred lecture mode, length, and assessment method as defaults unless the educator has a strong pedagogical reason or prior experience indicating that a different approach would be more suitable. Teaching students in their preferred mode and using the assessment methods they enjoy most—or avoiding those they enjoy least—could be advantageous. This is particularly important in the context of higher education, where educators (and institutions more generally) are often evaluated based on student satisfaction scores.

One limitation of this paper is that students are asked about their preferences for a generic module. It is possible that students might prefer online or longer lectures for some types of modules. Future research investigating students’ preferences over lecture mode and length for different types of modules would be valuable.

Recall that participation in both online surveys was incentivised with a £5 Amazon voucher, but attrition rates were relatively high (22% and 24% respectively). While these rates are comparable to other studies (e.g., Daikeler et al. [9], Liu and Wronski [23]), there are concerns that this may impact the generalisability of our findings (i.e., if students who drop out of the survey have different preferences than those who remain). Since are results hold across an entire cohort of master’s students (where we have no attrition) and all the other groups we consider, we argue that our findings are unlikely to be driven by selection. Therefore, these results are likely to reflect the preferences of students at their respective institutions (a top ranked business school and a mid-ranked research-intensive university). However, we fully acknowledge that these results might not hold across all programs and/or institutions. With that in mind, these results should be interpreted as suggestive evidence that the increasing use of online teaching may not necessarily align with students’ preferences. Future research investigating preferences across different programmes and institutions seems like an important next step.

A final limitation in this paper we measure expected learning and enjoyment on a 0–100 scale. This approach allows us to identify how different factors influence students’ expectations of leaning from and enjoyment of a module. However, it is not immediately clear what movements along these scales mean in terms of expected learning. Future research investigating how different aspects of module design influence students’ expected marks and/or job prospects seems like a natural next step.