Does class attendance still matter?

Abstract

This paper presents a study on the impact of class attendance on academic performance in a second year Electronics Engineering course module with online notes and no mandatory class attendance policy. The study shows that class attendance is highly correlated to academic performance, despite the availability of online class notes. In addition, there is significant correlation between class attendance and non-class contact with the lecturer and between student performance in the first year of university study and current academic performance and class attendance. However, there is no correlation between pre-university academic performance and current class attendance and academic performance. The study finds no gender bias in either class attendance or academic performance. Lastly, this being a course module open to students following different degree programmes, the study finds that a student’s choice of degree programme has no impact on class attendance and academic performance in this particular course module.

Introduction

Traditionally, lectures and tutorials have been the dominant forms of instruction in conventional face-to-face undergraduate courses. Research on class attendance has established that, on average, students with high attendance achieve higher academic performance in both coursework and examinations than students with poor attendance (Woodfield et al., 2006; Romer, 1993; Van Blerkom, 1992; Park and Kerr, 1990). With the advent of the internet, the provision of online lecture notes to undergraduate students has become the norm. However, to date little or no research has been carried out to establish the relationship between academic performance and class attendance in courses with online lecture notes.

The internet now gives students access to tailor-made online notes as well as external sources of information within and outside of their own institution. It is therefore important to establish the significance of class attendance to academic performance in this networked environment. For instance, if it is established that class attendance is no longer a significant factor in academic performance then the traditional lecture and tutorial method would need to be reviewed. At the very least this would entail adapting traditional undergraduate teaching practices to ensure that they remain relevant in today’s networked world; at most, the introduction of completely new approaches to undergraduate teaching which could lead to a complete redefinition of the whole concept of university education. The work reported in this paper is an initial attempt at establishing the relevance of class attendance at university in today’s networked world.

The statistical investigation context: the course module

The course module in question is a second year module on Communication and Networking Technologies which is offered in the second semester at the University of Exeter. This course module comprises 20 lecture periods of one-hour each. Lecture attendance is encouraged, but not mandatory. In line with university guidelines on best practice, online lecture notes are made available to the students prior to the lecture.

The Communication and Networking Technologies module is compulsory to Electronic Engineering students and optional to Computer Science and Information Technology and Management of Business (ITMB) students. For the academic year 2008/09, the Electronics Engineering cohort comprised nine students and both the Computer Science and ITMB cohorts each comprised 17 students.

All three cohorts were predominantly male dominated. There were no female students in either the Computer Science or Electronics cohorts and only six of the 17 students in the ITMB cohort were female. This male dominance is a feature of higher education courses in science, technology, engineering and mathematics in most industrialised countries, including the United Kingdom (Blickenstaff, 2005).

Assessment of the module is carried out through two tutor-marked assignments (TMA1 and TMA2), each worth 10% of the module mark, one Network Design and Simulation project worth 20% of the module mark and a two-hour, closed-book/note examination at the end of the course worth 60% of the module mark. The two tutor-marked assignments and the project are both summative and formative, with students required to explore and investigate subject material that goes beyond that covered by the lecture notes. To help students in this regard, each assessed piece of work is accompanied by a self-assessed assignment which is broadly similar to the tutor-marked assignment. Whilst no assistance is offered on the tutor-marked assignment, the lecturer is available, both inside and outside of class, to help students work through the self-assessed assignments, and detailed solutions to these self-assessed assignments are provided on request. Conversely, the final examination is based entirely on material covered in the lectures and made available in the online lecture notes.

To aid the learning process, material covered in lectures and posted on the web resources is closely synchronised with the work which needs to be covered by each of the tutor-marked assignments. For instance, time is set aside at the beginning of each lecture to address any queries associated with the current piece of assessment. In addition, following the introduction of new material, in-class exercises are used to expand on the lecture material and to shed light on the tasks to be carried out in each assignment. Collaboration is actively encouraged during these in-class exercises. Consequently, lecture sessions provide an avenue for students to interact both with each other and the lecturer.

Overview of class attendance over the whole semester

Figure 1 shows the percentage attendance for each lecture. The highest class attendance is 73%, which coincides with the release of the first tutor-marked assignment. The average lecture attendance over the whole semester was 56%, with significant dips for lectures 13 and 19 and a trend line with a small negative gradient of 0.0074. The attendance dips coincide with the submission deadlines for coursework in other modules. Previously, van Blerkom, in a study of class attendance patterns in undergraduate psychology classes (1992), also noted that class attendance decreased as the semester progressed and, most significantly, when students had to submit other coursework.

Figure 1 ECM2117 Percentage lecture attendances over the semester

Effect of class attendance on academic performance

Table 1 shows the correlation values between lecture attendance and academic performance. The associated p-values were 0.0022 for the correlation between project marks and attendance and less than 0.0001 for attendance and performance in the other assessment pieces, thereby confirming the statistical significance of all the correlation values in the table.

The correlation between lecture attendance and performance varied with the type of assessment in question. At 0.6884, performance in the exam had the highest correlation with attendance, followed by TMA2 at 0.6530, then TMA1 at 0.5927 and the project at a correlation value of 0.4601. Since the correlation between performance and attendance is positive in all the pieces of assessment it implies that attendance affects student performance. However, since this correlation is less than one in all types of assessment it follows that other factors also affect performance as well.

Given these different correlation values between performance in different types of assessment and attendance, a logical step would be to establish whether these differences are indeed statistically significant. Using Williams T2 formula (Cramer, 1994 p. 227), performance in each of the three pieces of coursework was compared to performance in the final examination. As shown in Table 2, the effect of class attendance on exam performance is significantly different from its effect on performance in both the Project and TMA1 but not in TMA2. Hence, on the basis of the data available, it cannot be concluded with certainty that class attendance affects differently a student’s performance in the exam and coursework.

Significantly, this study shows that class attendance still remains a key determinant of academic performance, even when the students have unfettered access to online lecture notes. Previous research (Vandehey et al., 2005; Grabe and Christopherson, 2005) has also found that the availability of online lecture notes does not in itself determine academic performance. Vandehey et al. (2005) observed that, when assessed by a common set of examinations, there was no significant difference in academic performance between students who were given complete notes, students who received outline notes and students who did not get any notes. Also, Grabe and Christopherson (2005) concluded that there was little or no correlation between note use and examination performance.

Both studies attribute the poor association between examination performance and access to online notes to study habits. In the study by Vandehey et al. (2005) at least 30% of the students with access to notes admitted to not making use of them at all. In the study by Grabe and Christopherson (2005) it was observed that students with low class attendance tended to use the available online notes only when they had to sit the examination. This suggests that class attendance may serve as a proxy for student engagement with the course module (Marks, 2000; Fredricks et al., 2004), with higher attendance suggesting higher engagement. Woodfield et al. (2006) have suggested that high absenteeism levels may be explained by a comparative lack of application on the part of the absentee.

Table 1 Correlation between attendance and performance

	Attendance	Exam mark	TMA1 mark	TMA2 mark
Exam mark	0.6884
TMA1 mark	0.5927	0.8343
TMA2 mark	0.6530	0.7564	0.7482
Project mark	0.4601	0.6454	0.6238	0.6534

Effect of previous performance on current performance

Tsai and Perry (1975) carried out a study to identify factors affecting students’ academic performance and persistence at a university in the USA. Academic performance was evaluated using college grade point average (GPA) whilst persistence was categorised by whether or not the student in question had previously withdrawn from other university studies. They found that a student’s performance in high school (secondary school) gives the best indicator for the student’s subsequent performance at university. Similarly, in a study of academic performance in an undergraduate money and banking course module, Park and Kerr (1990) found that cumulative GPA is a good predictor for the grades attained by the students on the course.

An investigation into the impact of pre-university performance on attendance and performance in the Communication and Networking Technologies course module has been carried out. Using the tariff table provided by the Universities and Colleges Admissions Service (UCAS, 2009), the UCAS points attained by each student prior to entering university were computed. The correlation between UCAS points and attendance in the Communication and Networking Technologies course module was 0.1073 at a p-value of 0.5271, whilst the correlation between UCAS points and student performance in the module was 0.0690 at a p-value of 0.6849. These statistical values suggest that there is no correlation between UCAS points and student attendance or performance in the Communication and Networking Technologies course module, contradicting the findings by Tsai and Perry. However, noting that the University of Exeter is highly selective (with typical offers only being extended to students attaining A-level grades ranging from AAB to BBB (University of Exeter, 2009)), the findings in this study suggest that the disparity in entry level performance for students embarking on Engineering, Computing and IT degree programmes at the University of Exeter is too small for it to be a reliable predictor of future student performance on the degree programme.

Table 3 shows that, unlike pre-university performance, academic performance in the first year of university study (stage one performance) correlates well to attendance and performance in the second year Communication and Networking Technologies course module. This is in agreement with McKenzie and Gow (2004) who, following a comparative study of the first year academic achievement of school leavers and mature students, concluded that for both groups of students previous university performance is a much stronger predictor of subsequent university performance than pre-university performance is of performance in the first year of university. McKenzie and Gow suggest that this may be due to differences between preuniversity and university study environments, as indicated by the fact that students are likely to experience more external regulation during their pre-university studies than at university where they are expected to direct their own learning.

Table 2 Williams T2 evaluation of the significance of performance differences in coursework and the final examination (degree of freedom = 29)

Coursework compared to the final examination	Williams T2 Statistics	Is coursework and exam performance difference significant?
Computed value	Critical value
TMA1	2.3089	1.6853	Yes
TMA2	1.187	1.6853	No
Project	1.896	1.6853	Yes

Table 3 First year performance correlation to current attendance and performance

	Stage one performance	Attendance
Attendance	0.6054	1.0000
ECM2117 course overall performance	0.5556	0.7788

Relationship between academic performance in course and stage average

The findings in the previous section seem to suggest that performance in a previous year of university study, unlike pre-university performance, is a strong predictor of subsequent university performance. It may therefore be reasonable to expect that a student’s performance in a given course module may serve as a predictor of the student’s overall performance at the degree stage in question. In this study it was observed that the marks attained in the Communication and Networking Technologies module are highly correlated to the stage average marks, with a correlation coefficient value of 0.8951 at a p-value of 0.0001. In addition, there is no statistically significant difference between the averages of the two sets of marks as evidenced by a Student t-statistics value of 0.068676 at the 5% significance level, which is much less than the critical t-value of 1.9908 for 78 degrees of freedom.

In a related study, Ma (2001) investigated the stability of academic performance across subject areas in elementary school and noted that there is significant correlation in student performance across individual subjects. It may therefore be concluded that a student’s performance in individual course modules is a significant predictor for the student’s overall academic performance.

Table 5 shows that, at 67%, the Electronics Engineering cohort had the highest mean attendance, whilst the Computer Science cohort had the lowest mean (54%). However, as the student t-statistical data in Table 6 shows, there are no statistically significant differences in the mean cohort attendances (at a significance level of 5%). This suggests that attendance is not affected by the degree programme being followed by the student.

Table 7 shows that the Electronics Engineering cohort had the highest mean in the final examination and the ITMB cohort had the lowest. The ITMB cohort had the highest mean in each of the three courseworks and the Computer Science cohort had the least mean in two of the courseworks. However, the student t-statistics values shown in Table 8 suggest that, at the 5% significance level, there is no statistically significant difference in the mean cohort mark. This implies that performance in the course module is not affected by the degree programme being followed by the student.

Table 4 Statistical comparison of ECM2117 course marks and stage average marks

Statistical value	ECM 2117 marks	Second year stage average marks
Mean	57.625	57.358
Variance	425.983	177.499

Table 5 Statistics for attendance by each of the three cohorts

Degree programme	Number of students	Mean attendance (%)	Attendance variance
ITMB	17	59	1099.265
Computer Science	17	54	1127.390
Electronics Engineering	9	67	1313.194

Table 6 T-statistics for the statistical differences in mean attendance by cohort at a significance level of 5%

Degree programmes being compared	t-statistics
Computed value	Critical value
ITMB vs Computer Science (32 degrees of freedom)	0.4369	2.0369
ITMB vs Electronics Engineering (24 degrees of freedom)	0.5672	2.0639
Computer Science vs Electronics Engineering(24 degrees of freedom)	0.9144	2.0639

Table 7 Statistics for performance by each of the three cohorts

Degree programme	Final examination mark	TMA1 mark	TMA2 mark	Project mark
Mean (%)	Variance	Mean (%)	Variance	Mean (%)	Variance	Mean (%)	Variance
ITMB	49.94	669.028	68.53	763.639	68.76	717.567	68.06	861.433
Computer Science	57.21	276.252	62.53	643.390	56.76	786.191	61.71	527.971
Electronics Engineering	63.94	534.090	59.78	1232.444	63.444	768.278	64	1012.250

Table 8 T-statistics for the statistical differences in cohort mean mark at a significance level of 5%

Degree programmes being compared	t-statistics
ITMB vs Computer Science (degrees of freedom: 32; t-critical = 2.0369)	ITMB vs Electronics Engineering (degrees of freedom: 24; t-critical = 2.0639)	Computer Science vs Electronics Engineering (degrees of freedom: 24; t-critical = 2.0639)
TMA1	0.6595	0.7000	0.2303
TMA2	1.2759	0.5801	0.4762
Project	0.7027	0.3261	0.2120
Final exam	0.9744	1.3595	0.8584

Effect of gender on class attendance and academic performance

Gender bias in both attendance and academic performance was investigated using student t-statistics. As shown in Table 9, whilst female students appear to have a higher mean attendance than male students (71% versus 52%), this is not statistically significant, as indicated by the computed t-statistics value of 1.1397 which falls short of the critical value of 2.1314 for 15 degrees of freedom for a significance level of 5%. However, these results are in contrast to those obtained in a more comprehensive study involving 650 undergraduate students at the University of Sussex which found a statistically significant gender bias in class attendance in favour of female students (Woodfield et al., 2006). In the light of the University of Sussex study it appears that a reliable and representative assessment of gender bias would require a larger student population than the one assessed in this study.

As shown in Table 10, female ITMB students consistently outperform male ITMB students in the three continuous assessments (TMA1, TMA2 and the project), with the male students outperforming the female students in the final examination. However, the computed t-statistics value for all pieces of assessment falls short of the critical value of 2.1314 for 15 degrees of freedom (Table 11). This suggests that gender bias in performance in both the final examination and the courseworks is not statistically significant.

Relationship between class attendance and lecturer contact

Previous research indicates that student non-class contact with lecturers is an important predictor of academic performance. For instance, Pascarella et al. (1978) analysed freshman GPA and observed that the high student-faculty non-class interaction was associated with high student motivation and academic performance, whilst Chickering and Gamson (1987) noted that frequent interaction between faculty and students, both in and out of classes, improves student motivation and involvement. In the same vein, Lundberg and Schreiner (2004) determined that student-faculty non-class interaction contributes as much as one quarter of the student’s total learning.

In this study, students could access the lecturer outside the class via email or by visiting him in his office. More often than not, visits to the lecturer’s office were arranged through email. Hence, email records can be assumed to contain sufficient evidence of the extent to which each student interacted with the lecturer outside class. As shown in Table 12, those students who contacted the lecturer at least once by email had a higher mean attendance than those who never contacted the lecturer (69% versus 48%). This difference is statistically significant, as evidenced by the student t-statistics value of 2.0795 which exceeds the critical value of 2.0211 for 40 degrees of freedom at the 5% significance level.

Table 9 Statistics for ITMB attendance by gender

Gender	Number of students	Mean attendance (%)	Attendance variance
Male students	11	52	1376.363
Female students	6	71	484.167
t-statistics for difference in gender attendance: [degrees of Freedom: 15 t-critical (one-tailed) = 2.1314]	1.1397

Table 10 Statistics for ITMB performance by gender

Gender	Final examination mark	TMA1 mark	TMA2 mark	Project mark
Mean(%)	Variance	Mean(%)	Variance	Mean(%)	Variance	Mean(%)	Variance
Male students	53.27	782.618	65.18	1137.564	63.36	1025.655	66.91	1154.091
Female students	43.83	506.467	74.67	98.667	78.67	63.067	70.17	440.167

With regard to performance, as shown in Table 13, students who interacted with the lecturer out of class performed better in all assessments than those who did not. This difference in performance is statistically significant, as evidenced by the fact that all the student t-statistics values for all assessments exceed the critical value of 2.0211 for 40 degrees of freedom at the 5% significance level (Table 14).

The findings regarding class attendance and student-lecturer non-class contact can be summed up thus: a high level of class attendance is associated with a high degree of non-class student-lecturer interaction and high academic performance. Furlong et al. (2003) note that students’ perceptions of successful interaction with their teacher increase when the teacher is seen to be supportive, and this leads to greater student academic motivation and improved attitudes towards school. Hence it may be possible that students with a high level of class attendance perceive the lecturer as supportive and are therefore likely to have a higher level of non-class student-lecturer interaction when compared to students with lower attendance. As Astin (1984) observes, students who interact frequently with faculty members are more likely than other students to express satisfaction with all aspects of their institutional experience, including student friendships, variety of courses, intellectual environment and even the administration of the institution.

Summary of study findings and their implications

The main aim of this study was to determine whether or not class attendance had an effect on student academic performance in a second year course module with online notes and no mandatory class attendance policy. The study also sought to establish the factors associated with high and low class attendance. Possible factors investigated included pre-university performance (as indicated by the students’ UCAS entry points), performance in the first year of study, choice of degree programme, gender and student-lecturer non-class interaction.

Table 11 T-statistics for differences in gender performance at a significance level of 5%

Assessment	t-statistics (degrees of freedom: 15; t-critical = 2.1314)
TMA1	0.6644
TMA2	1.1358
Project	0.2121
Final exam	0.7078

Table 12 Comparison of attendance figures for students who kept contact with the lecturer and those who did not

Students	Number of students	Mean attendance (%)	Attendance variance
Students with at least one contact with lecturer	21	69	1408.690
Students who never contacted lecturer	21	48	732.857
t-statistics for difference in attendance [degrees of freedom: 40; t-critical = 2.0211]	2.0795

The study found that class attendance is a key determinant for academic performance in courses with online lecture notes. In addition, there is strong correlation between class attendance and non-class contact with the lecturer. The study finds no correlation between pre-university and current academic performance or class attendance. However, for the second year module analysed in this paper, there was a high correlation between a student’s current level of class attendance and academic performance and the student’s academic performance in the first year of university study. Unlike previous research, this paper could not confirm that there is gender bias with regard to class attendance and academic performance. In addition, no statistically significant evidence could be found to assert that class attendance and academic performance in the course discussed in this paper were in any way affected by the degree programme for which the student was studying.

The main implication arising from this study is that, since class attendance is so clearly associated with academic performance, lecturers ought to improve and maintain high class attendance rates. A simplistic approach to this would be to put in place a mandatory attendance policy or, where this is not feasible, to use appropriate incentives to encourage students to attend classes. However, as Rodgers (2002) found out, simply increasing attendance does not necessarily lead to improved academic performance. This therefore means that a better approach to achieve meaningful attendance may be to focus on issues that enhance student engagement.

The term student engagement is used to refer to the extent in which the student is involved with his/her studies. Marks (2000) characterises student engagement as ‘a psychological process, specifically, the attention, interest, and investment and effort students expend in the work of learning’. Fredricks et al. (2004) have categorised engagement into behavioural, cognitive and emotional engagement. Behavioural engagement entails positive conduct, such as following rules and avoiding disruptive behaviours such as skipping school and getting into trouble. In addition, behavioural engagement is also concerned with involvement in learning and academic tasks and includes conduct such as effort, persistence, concentration, attention, asking questions and contributing to class discussion. Emotional engagement refers to students’ affective reactions in the classroom, including interest, boredom, happiness, sadness and anxiety. Cognitive engagement focuses on the psychological investment in learning, a desire to go beyond the required level and a preference for challenge.

Table 13 Statistics for comparing performance of students who kept contact with the lecturer and those who did not

Students	Final examination mark	TMA1 mark	TMA2 mark	Project mark
Mean(%)	Variance	Mean(%)	Variance	Mean(%)	Variance	Mean(%)	Variance
In-contact students	65.714	469.889	75.952	112.648	73.095y	1025.655	74.667y	1058.714
Non-contact students	46.952	344.523	52.048	1223.448	51.667	229.790	54.714	266.133

Table 14 T-statistics for differences in performance of students who kept contact with the lecturer and those who did not at a significance level of 5%

Assessment	t-statistics (degrees of freedom: 40; t-critical = 2.0211)
TMA1	2.9968
TMA2	2.7176
Project	2.5121
Final exam	3.0128

How then can the lecturer improve student engagement with a view to improving meaningful attendance? The findings on the association between class attendance, student-lecturer non-class interaction and academic performance would suggest that the process of learning and teaching should be viewed as a socio-cognitive phenomenon that requires a supportive social environment in order to be effective. Battistich et al. (1993) report that supportive environments in which members are friendly, help one another, show concern for one another’s welfare and work collaboratively are associated with increased liking for school, greater intrinsic motivation, concern for others and self-esteem. The lecturer could therefore improve class attendance by fostering a community sense within the classroom characterised by supportive student-lecturer relationships, mutual respect between the lecturer and students, and between students themselves, and adopting a cooperative pedagogic approach in which learning is viewed as a shared activity between all members of the community, including the lecturer (Furlong et al., 2003).

Limitations of current study and suggested improvements

The main limitations of this study are that it has focussed on a single course module with a relatively small class size of 43 students taught entirely by one lecturer. The findings obtained would be more representative if class attendance was investigated across all the course modules being undertaken in the school and across all stages of the undergraduate programmes. Another limitation of this study is that no interviews were conducted with the students, thus the student opinion has not been canvassed. Lastly, as suggested by Furlong et al. (2003), the school environment has an effect on the way in which students perceive learning. Therefore, for the findings to be truly representative, similar studies must be carried out in Computer Science and Engineering departments in other universities. The study is being expanded to take account of these limitations.