The relationship between attendance and academic performance in higher education has been explored extensively for decades. The debate on whether it is necessary to require mandatory attendance in secondary institutions has been going on concurrently. Most previous studies found a positive relationship between attendance and academic performance (Devadoss & Foltz,
1996; Kirby & McElroy,
2003). Romer (
1993) first found a significant positive correlation between performance and attendance based on an analysis of n = 195 undergraduate students’ attendance and course performance and advocated for mandatory attendance in school to promote performance. Durden and Ellis (
1995) then defined attendance as a proxy for motivation. They collected n = 346 students’ self-reported absence records, examined the relationship between attendance and academic achievement in an economics course, and found that absenteeism led to poor academic performance. Credé et al. (
2010) later conducted a systematic review on 69 empirical studies and found that lecture attendance was a significant medium-strong predictor of academic performance, before and after controlling for other potential confounding variables, such as student age, gender, grade, SAT score, IQ, hours of employment, and motivation levels. More recently, similar findings were also presented by studies across different subjects with varying effect sizes (Hollett et al.,
2020; Louis et al.,
2016). For example, Landin and Pérez (
2015) recruited four cohorts of university students from a pharmacy course and correlated their attendance with performance separately. Positive correlations were observed across all four cohorts, suggesting a positive effect of attendance on performance.
Andrietti (
2014) also analyzed longitudinal data from undergraduate students enrolled in an introductory macroeconomic course across the academic year to evaluate the relationship between lecture attendance and academic performance using proxy variable regressions. Findings revealed that attendance had a moderate positive impact on performance, although the effect disappeared after introducing time-invariant variables. This suggests that unobservable mechanisms such as students’ characteristics or motivation may interact with the relationship between attendance and performance. Similarly, Krohn and O’Connor (
2005) observed students in three undergraduate macroeconomics courses and found a positive significant effect of attendance. However, the relationship became non-significant when instrumental variable techniques were applied to analyze the data collected during the term.
No relationship as well as minimal or conditional relationships between attendance and performance have increasingly been found in recent studies (Andrietti & Velasco,
2015; Büchele,
2020; Choi-Lundberg et al.,
2020; van Walbeek,
2004) and were attributed to two main reasons. First, with the world-wide digitalization of education, students no longer must attend classes to gain access to course materials, so attendance is not vital for achievement (Büchele,
2020). Second, unlike previous studies that only correlate performance with attendance, more studies seek to address the endogenous bias of attendance by controlling confounding variables, such as student characteristics and motivations (Choi-Lundberg et al.,
2020), or introducing mediators that are related to engagement, such as task engagement, tutorial engagement, and metacognition regulation (Büchele,
2020).
On the other hand, Schneider and Preckel (
2017) argued that the effect of attendance on learning outcomes has remained significant and withstood the great advance of learning technologies over the years. They conducted a systematic review of 38 meta-analyses to investigate the variables associated with achievement in higher education. Class attendance (
d = .98, ranked 6) ranked the sixth most significant predictor for academic achievement among all the 105 variables examined, and ranked the most significant predictor within student variable category. In addition, their study revealed that online courses and blended courses does not seem to mitigate the importance of class attendance for academic achievement. However, they argued that it is still too early to draw conclusions on mandatory attendance policies before the mechanism underlying class attendance has been fully understood when information and educational technology overtake the field of education.
Undoubtedly, attendance has been proved to impact performance. However, there are still some unresolved issues that remain to be further studied on this topic. First, most previous studies adopted self-reported attendance records as predictors of academic achievement, in which researchers requested participants to recall their attendance rate at the end of the semester. The self-reported attendance rate introduces measurement bias. Second, attendance is an endogenous factor for learning, with highly motivated and high-achieving students being more likely to attend lectures regularly and engage in the class contents, and thus, achieve higher course performance (Andrietti,
2014). Although some studies attempted to control student-level variables to mitigate the upward endogeneity error of attendance, few incorporated the instruction-level variables, such as in-class activity engagement, peer or teacher interaction, or performance on formative assessments. The potential measurement error and endogeneity bias may severely attenuate the validity of the conclusions presented in the related research.
Based on the TEL engagement, in-class attendance serves as an indicator of traditional school engagement. Concomitantly, online engagement is indicated by self-regulated online learning activities and performance on online formative assessments. Both traditional and online engagement may be essential determinants of academic success. In the era of TEL-based education, more research needs to be done to understand the dynamics among in-class engagement, online engagement, and academic performance. Moreover, the potential mediating effects of TEL engagement indicators upon the relationship between in-class engagement and academic performance are also underexplored.
Technology-enhanced learning has become a major trend in education, especially in today’s climate of the COVID-10 pandemic. TEL transforms the conditions of engagement learning from traditional classroom-based to blended and, currently, to fully online through various digital technologies (López-Pérez et al.,
2011; Nouri et al.,
2016). A substantial body of literature has investigated the relationships between student online self-regulated learning and self-assessment with academic performance using data extracted from LMSs (e.g., Hung & Crooks,
2009; Shi et al.,
2015). Most previous studies have concluded consistent results with traditional schooling contexts that higher levels of TEL engagement could facilitate academic success (Hung & Crooks,
2009; Kibble,
2007; Zacharis,
2015). However, few have investigated the associations among traditional engagement indicators such as attendance, TEL engagement indicators such as self-regulated learning and self-assessment, and academic performance.
As an essential TEL engagement indicator, students’ online self-regulated learning plays an increasingly important role in the formal contexts of higher education, for LMSs such as Canvas, D2LBrightspace, Moodle, and Sakai have been regarded as critical digital tools that assists faculty members in delivering poly-synchronous materials, lectures, and assessments (Gautreau,
2011; Washington,
2019). Some of the studies have been done to evaluate the relationship between attendance, online learning engagement, and performance in online learning in higher education (Bekkering & Ward,
2020; Doggrell,
2020; Nieuwoudt,
2020). Doggrell (
2020) inspected the associations between lecture attendance, lecture recordings access, and academic achievements on n = 117 medical students sampled from two sessions of medical laboratory science courses. They found that, with the availability of lecture recording, there is no significant correlation between lecture attendance and academic achievement. They suggested that using a mixture of multimedia educational technologies is likely to ensure higher academic success.
Online formative assessment is another important indicator of TEL engagement that predicts performance (Gikandi et al.,
2011; Spector et al.,
2016). Educators need to consider formative practices and optimally integrate them into their teaching and assessments. Online formative assessment also provides learners with self-evaluation and feedback to help them orient and adapt their own self-regulated learning (Zimmerman,
2002). Gikandi et al. (
2011) conducted a review of literature on 19 empirical studies about online formative assessment in the context of online learning in higher education. They found that online formative assessment effectively promoted learner engagement and learner community development. Other studies also confirm the constructive, beneficial effect of formative assessment on learning outcomes (Rakoczy et al.,
2019; Robinson & Udall,
2006).
With the fast development of the areas of educational data mining (EDM) and learning analytics (LA), a great number of studies emerged using EDM and LA to measure online engagement and learning by analyzing web-based log event data generated during the LMS usage recording the users’ activities, IP address, date, and time sequence (Aldowah et al.,
2019; Dutt et al.,
2017; Papamitsiou & Economides,
2014; Romero & Ventura,
2010; Romero et al.,
2008). Common practices of EDM and LA include applying feature-engineering techniques to extract engagement indicators, such as analyzing the text posted on online forums (Larsen et al.,
2008) or counting individuals’ click frequencies and total time spent in different LMS sessions throughout a course (Geigle & Zhai,
2017; Zacharis,
2015). Students’ online learning engagement can be objectively reflected by their actual web usage on the LMS. Most studies conducted using EDM/LA approaches reported that higher levels of self-regulated learning are positively correlated with academic performance (Geigle & Zhai,
2017; Hung & Crooks,
2009; Zacharis,
2015). However, few studies have explored the impact of online learning engagement as captured by features extracted from log data on the relationship between in-class lecture attendance and academic performance.
Prior knowledge is constantly regarded as a significant student characteristic to predict performance in TEL education (Asarta & Schmidt,
2017; Kinsella et al.,
2017; Schneider & Preckel,
2017; Song et al.,
2016; Spires & Donley,
1998; Tobias,
1994). Prior knowledge is defined as the information or experiences that a learner already established regarding a new topic either taught from learning or drawn from experiences (Tobias,
1994).
Previous studies commonly found that prior knowledge is positively related with academic performance through the facilitation of higher levels of motivation, engagement, and self-regulation (e.g., Schneider & Preckel,
2017; Song et al.,
2016). Song et al. (
2016) conducted a study to examine the effects of prior knowledge, self-regulation, and motivation on performance via structural equation modeling. They assessed 386 medical clerk students’ prior knowledge through multiple choice items and measured their self-reported self-regulation and motivation. A knowledge post-test and a clinical reasoning test were administered as performance measures. Findings revealed both direct and indirect positive correlations of prior knowledge with learning outcome and self-efficacy. Conversely, students with little or no prior knowledge will be disadvantaged when they process and memorize entirely new information. In the worst case, students with false prior knowledge will have to correct and update the false information and reconstruct their knowledge system (Kowalski & Taylor,
2009). From a systematic review of meta-analyses of variables associated with achievement in higher education, Schneider and Preckel (
2017) also found that prior intelligence or prior knowledge is an important predictor for achievement (d = .90, ranked 7 out of 105).
Given the fact that prior knowledge is reported to account for a large proportion of variances of learning outcomes (Schneider & Preckel,
2017; Song et al.,
2016; Tobias,
1994), the present study controlled the effect of prior knowledge when examining the relationship among attendance, self-regulated learning, performance of formative assessments, and academic performance to exclude the confounding bias.
Gaps identified in the previous studies
With the rapid digitalization of education around the world, online learning and formative assessment have become essential components of both formal and informal learning in higher education and the key to academic success. The findings on impact of attendance on performance are no longer valid if online learning and online formative assessment are not considered and evaluated.
Moreover, most previous studies adopted an instrumental approach, such as the National Survey of Student Engagement (NSSE: Ewell,
2010; Kuh,
2009), the Australian Survey of Student Engagement (AUSSE: Coates,
2010), or the Utrecht Work Engagement Scale for Students (UWES-S: Carmona-Halty et al.,
2019; Seppälä et al.,
2009) to measure engagement and other student characteristics. The self-reported scales exhibit inherent measurement errors and may not reflect students’ real level of engagement.
The advancement of educational data mining (EDM) and learning analytics (LA) methods could provide more insights in TEL contexts. The LMS can be used to record attendance more accurately, compared with the self-reported attendance rate recalled by the students. In addition, researchers can extract students’ online learning activities from the automated generated log files in the LMS. The use of LMS also enables instructors to examine the students’ prior knowledge, to organize in-class online activities, to administer online formative assessments in and outside the classroom, and to revise their instruction because of the way students interact with the materials. The collection of all the information above through LMS and its inclusion into the analysis of attendance and performance can help minimize the endogeneity and measurement bias mentioned in previous studies.
Thus, we propose a learning analytics approach to measure students’ lecture attendance and online learning engagement through information extracted from the log file generated by LMS. Additionally, we regard TEL engagement indicators—self-regulated online learning and online formative assessment administered on LMS—as important indicators of academic performance in addition to traditional engagement indicator attendance.