Synchronous online courses gained popularity during the pandemic and have since become the predominant course delivery format for online courses due to their ability to reduce educational cost while preserving real-time communication and immediate feedback (Bailenson, 2021). The rapid growth of synchronous course prompts questions about improving the quality of online courses at scale (Bettinger et al., 2017; Lowenthal et al., 2019; Russell & Curtis, 2013; Xu & Xu, 2020).. One area of intense debate in this context is the role of class size in online courses. However, limited research has quantitatively examined the effects of class size on student learning outcomes in college synchronous online courses. Assessing these effects is a pertinent issue: If increasing class size in synchronous online courses does not compromise student learning outcomes, it opens up the possibility for departments to consider offering larger synchronous online classes. This approach could help reduce educational costs and enhance accessibility without sacrificing student engagement.

This study addresses this gap by answering three main research questions. First, what is the impact of class size on students’ academic performance and course satisfaction? Second, what are the mechanisms through which class size effects operate? Third, how do class size effects vary along the distribution of class sizes (Non-linear effects)?

We analyzed data from an anonymous research university (ARU hereafter) in China. Due to the COVID-19 pandemic, all courses in the spring semester of 2020 in ARU were delivered online synchronously, with approximately 82% delivered via ClassIn, which was a popular platform for online course delivery in China at the time. This study focused on undergraduate students enrolled in synchronous online courses for the duration of the semester. Furthermore, this study examined courses categorized as “lecture” sections, excluding physical education and lab courses. As a result, our sample comprised 6,603 undergraduate students enrolled in one of the 638 synchronous online classes offered by 30 departments. We obtained data from two sources: (1) administrative data, which includes students’ and instructors’ demographic characteristics, class enrollment size, students’ academic performance, and instructors’ teaching evaluation, etc., and (2) clickstream data generated by the ClassIn platform, which captured information such as the length of time students were assigned to interact with their peers and instructors, as well as their time spent in the virtual classroom.

We began by estimating the effects of class size on students’ academic outcomes and course satisfaction. Our analyses indicate that class sizes negatively affect course grades and course satisfaction in synchronous online courses. Drawing on the rich clickstream information generated by the online platform, we examined two channels through which class size effects may operate: (i) students’ course attendance and (iii) course interaction opportunities. Our findings suggest that reduced course interaction opportunity is the most robust channel through which larger classes negatively affect students’ academic outcomes and course satisfaction. In addition, we explore non-linearities in the class size effect and heterogeneity by students’ academic preparation, grade level, and course credits. Our findings indicate a consistent negative effect across the entire spectrum of class sizes, with larger class sizes exhibiting increasingly detrimental effects. We also found that the negative relationship between class size and student outcomes is highly robust across different types of students.

We analyzed the effects of class size on students’ outcomes, using within-student variation in class sizes: yicm= αi + βm + γCScm + δXcm + θPcm + φTcm + εicm (1) where yicm is an outcome, such as course grade for student i in class c in department m. αi represents student fixed effect, which allows for comparisons among different classes taken by the same student, thereby mitigating bias associated with students’ self-section into different classes. βm represents department fixed effects, enabling comparisons among different classes within the same field of study. CScm captures the class size in class c of department m, defined by the number of students enrolled in the classes and the average number of students presenting in the class over the semester. Xcm represents control variables on class c in department m, such as course credits and course classifications. To account for the peer group composition within a class, we controlled for Pcm, including the share of male students, the proportion of students with average grade points in the lowest quartile, and the proportion of seniors. Tcm captures the characteristics of teachers in a course, such as their gender, job title, age, educational attainment, etc. However, considering that more than one faculty member can teach a class, the term Tcm represents the faculty composition of class c. This term includes the number of faculty involved in teaching the particular class, the proportion of male faculty, the proportion of professors, the ratio of overseas returning faculty, the average age of teachers, and the average teacher evaluation score. Finally, the error term εicm was clustered by course to capture common unobservable shocks to students’ outcome variables. We further calculated the implied effect size. The measure estimates the proportion of the within-student standard deviation in outcome variables that can be explained by a one standard deviation increase from the mean class size. We then examined whether there were any nonlinear class size effects using Equation (2). To do so, we categorized students into four quantiles based on the distributions of both class enrollment sizes and actual class sizes. yicm= αi + βm + ∑ γqCSqcm + δXcm + θPcm + φTcm + εicm (2) where CSqcm equals to one if the class size is in the qth quantile of class size distribution, and zero otherwise.

There is robust evidence of a negative class size effect on students’ academic achievement. Our study also revealed evidence of nonlinear class size effects on course grades in synchronous online courses. Specifically, we observed a negative impact between the first two quantiles and the last two quantiles. The results indicate that there were beneficial effects when moving both from mid-sized to smaller classes and from the very large to large classes. Our analysis also sheds light on the negative impact of class sizes on student course satisfaction. Unlike the nonlinear effects observed on course grades, the impact on course satisfaction showed a distinct pattern. It became more significant when moving from the first to the second quantile (class sizes ranging from 2 to 15 students and 16 to 24 students). However, there seems to be no further detrimental effect when moving from the second to the third quantile or from the third to the fourth quantile. Therefore, it appears that the class size range of 16 to 24 students was where the negative class size effect on course satisfaction reached its highest magnitude. In terms of the mechanisms, our findings suggest that, on average, class size did not have a significant association with class attendance. However, we did observe nonlinear effects where the course attendance rate began to decline when class sizes exceeded 24 students. Additionally, we consistently observed notable negative correlations between class size and student course interaction opportunities. In conclusion, our study highlighted the importance of considering class size as a factor influencing student attendance and course interaction opportunities.

Bettinger, E. P., Fox, L., Loeb, S., & Taylor, E. S. (2017). Virtual Classrooms: How Online College Courses Affect Student Success. American Economic Review, 107(9), 2855–2875. https://doi.org/10.1257/aer.20151193 Lowenthal, P. R., Nyland, R., Jung, E., Dunlap, J. C., & Kepka, J. (2019). Does class size matter?: An exploration into faculty perceptions of teaching high-enrollment online courses. American Journal of Distance Education, 33(3), 152–168. https://doi.org/10.1080/08923647.2019.1610262 Russell, V., & Curtis, W. (2013). Comparing a large- and small-scale online language course: An examination of teacher and learner perceptions. The Internet and Higher Education, 16, 1–13. https://doi.org/10.1016/j.iheduc.2012.07.002 Xu, D., & Xu, Y. (2020). The ambivalence about distance learning in higher education: Challenges, opportunities, and policy implications. In L. W. Perna (Ed.), Higher Education: Handbook of Theory and Research (Vol. 35, pp. 351–401). Springer International Publishing. https://doi.org/10.1007/978-3-030-31365-4_10