Non Attendance at Lectures: an Empirical Study

Non Attendance at Lectures: an empirical study

This is one of a set of papers and work in progress written by research postgraduates (MPhil and PhD) at Lancaster University's Department of Educational Research. The papers are primarily offered as examples of work that others at similar stages of their research careers can refer to and engage with

Abstract:

Empirical evidence is presented about the association between attendance at lectures and performance in assessed work using data gathered from a sample of over 400 students studying a first year undergraduate module taught at a business school. After controlling for various factors (age, ability, gender, mode of study between part and full-time, whether studying for an HND or degree, or live in College or private accommodation) the results suggest that attendance itself explains 24% of the variability in performance. Students in this sample who miss lectures saw their results decline by around 2.5% per lecture missed, which represented nearly a full grade classification between the average and full attending students on this module. Although attendance is only one of a number of influences on performance, the results of this study suggest strongly the importance that more empirical research is undertaken in this area, not necessarily because of what it says about the value of lectures, but also because of what it says about student motivation and effort.

Date:December 2001

Introduction

The purpose of this paper is to present empirical evidence of the relationship between how much a student attends lectures and how this attendance is associated with his/her performance in assessed work. The premise for the study is that there is value in attending which increases the learning achieved by the student, and that this learning is demonstrable in the marks achieved in assessed work.

While suggesting that those students who attend lectures more regularly are likely to perform better in assessed work may appear axiomatic, it does need a more rigorous analysis. In previous empirical on performance and lecture attendance considered later, the research of three US academics is often quoted as a means of explaining why those who attend lectures more should perform better. The first is Astin (1970) who argued that the degree of involvement of the student is crucial to the way s/he learns and therefore performs. Presented as an input/process/output conception of learning, attendance at lectures for Astin (1970) is construed as an input (involvement) from the student, which will increase output; one output being an improved level of performance in assessed work. The contribution from the second academic, Tinto (1987), was to recognize that this involvement of the student will be reflected in the quality of the relationships formed with the social and study aspects of university life. Students with positive social and study experiences are likely to stay in College (Tinto was primarily concerned with student retention) and by implication should attend classes more regularly and perform well. The third academic is Schmidt (1983). His interest was in the relationship between the way students allocate their study time and how this affects their performance. His analysis of 216 economics students at the University of Wisconsin showed that time spent in lectures was much more important in predicting performance in assessed work[1]. Students who spent the same amount of time studying but did not allocate sufficient time to lectures were, in effect, being irrational in the way they managed their time.

Although these three academics provide a rationale with evidence that helps explain why there is an association between lecture attendance and performance, there is a danger of misrepresenting their ideas. The most obvious concern is that attending lectures per se is not what determines performance. Lecture attendance may simply be a proxy for other factors which determine how a student performs. The more involved students in Astins (1970) model might, for example, be studying more diligently and performing better because of it, rather than from attending more lectures. The same is true of the students who have adjusted well to University life in Tintos (1987) framework and the evidence presented by Schmidt (1983) could be interpreted in another way. In assuming that an hour of study time was equivalent irrespective of the activity, Schmidt (1983) did not take account of possible learning efficiencies. Some of the students in his sample might have been working much more efficiently when working in groups, preparing for assignments and so on, and therefore taking less time to achieve their learning objectives when engaged in these study activities. For them, hours spent in lectures may be an exaggeration of their true learning and the importance of lecture time, over which they have no control, might simply have been exaggerated[2]. His results therefore might have been biased towards the importance of lecture attendance.

Empirical studies into the association between lecture attendance and performance has been dominated by academic economists in the US. A debate amongst some of them in 1994 (Bauer et al, 1994) demonstrated the range of opinion that coexists about whether attendance itself, or other latent influences for which attendance may be a proxy, are causally related to performance. The debate in effect revealed two schools of thought. There were those who argued that students are utility maximisers, that is they will attend class so long as it adds something to their overall satisfaction in some way. Attendance for them is a rational decision which takes account of performance. For these students, increasing attendance beyond their choice level would probably lower their utility, is unlikely to improve their performance and may in fact affect their performance deleteriously. The second view centred on the belief that there are obvious benefits to attending which as professionals we should be concerned about. Students do not necessarily realize what the benefits of attending are, and university tutors should therefore try and manage attendance in some way to improve student learning and performance[3].

This paper seeks to add to this area of study by presenting an empirical analysis of data on lecture attendance and performance for students at a UK University. By analyzing a database for a first year undergraduate module at a business school, results are presented which:

-measure the association between attending lectures and overall performance

-measure the association between attending lectures and performance on different types of assessment

-offer guidance on where further research might be needed in this area.

This is the first empirical study to look at student attendance and performance in assessed work for a UK University since the 1970`s[4].[5]

Empirical Evidence on the Association Between Lecture Attendance and Student Performance

A number of empirical studies were conducted in the 1990`s into whether there is an association statistically between lecture attendance and student performance. In each study the approach was to regress attendance with performance in assessed work, and to include control variables which seek to capture other potential determinants of performance. The specific contribution of attendance to performance was then identified separately alongside the other variables.

The most influential study was by Romer (1993). He looked at the attendance of students at three reputable universities and was surprised to find that `on a typical day at a typical elite American universityroughly one-third of the students in economics courses are not attending class` (p 168). He then undertook an analysis of attendance and performance of 195 students taking his macroeconomics course at the University of California, Berkeley. Although overall absenteeism was a little lower (25%) his analysis strongly supported the idea that attendance and performance are causally related. He presented his results in five linear regressions. The first was simply performance[6] (P) and attendance measured as a fraction of all the lectures attended (X):

P =1.25 + 2.19X

(0.27) (0.35)R-square = 0.31

 parentheses refer to standard error of the variable.

Attendance alone explained 31% of the variability in performance of these students and the coefficient on attendance was statistically significant. After controlling for student ability (Y), defined as the Grade Point Average of the student, and looking only at students who had done all problem sets on the course (a measure of motivation), the results for the fifth regression were:

P=-0.78 + 1.38X + 0.86Y

(0.43)(0.58) (0.14)R-square = 0.48

 parentheses refer to standard error of the variable.

In other words, attendance was still significant in predicting the performance of the more motivated students. He estimated that a student of similar ability and motivation could improve his/her grade from C+ to B+ by attending all, as opposed to a quarter, of lectures. He suggested the need to experiment with policies on mandatory attendance.

Romers` approach was developed in two further studies. Chan et al (1997) investigated the link between attendance and performance of 71 students studying Principles of Finance at a mid-western regional state university. They included a larger number of control variables than Romer, such as number of weekly hours worked, and an estimate of students perceived quantitative skills[7] and added to the Romer study in two ways. First, the data was adjusted by controlling for what they termed the `survival bias` in Romers work by including data for those students who had withdrawn from the course. They argued that all students should be included in the analysis including those who had not completed (survived) the course. The second addition was to use a control group for whom attendance was compulsory in order to respond to Romers suggestion about experimenting with making attendance mandatory. Their results confirmed largely those of Romer, `for every1% change in attendance, there will be 1.32% change in students courseaverage.` (p. 63). However, the performance of students in the control group did not improve sufficiently in order to justify attendance being made mandatory. The other follow up study to Romer was by Maloney & Lally (1998) of economics students at the National University of Ireland, Galway. This study focused on attendance and performance in examinations of 82 second and 121 third year students. Although the results supported the relationship between attendance and performance suggested by Romer, the coefficients and the variability explained were found to be lower which they suggested may be due to the contextual differences between students studying in the US and Ireland. An important outcome, however, was the difference in the relationship between second and third year students. Not only did second year students attend more regularly than third years (65% as opposed to 61%), the results suggested that attendance was much more important to second years than third years in determining performance. This was explained by the authors as third year students possibly having more experience and knowledge from the second year which enables them to be more informed about which lectures to miss. Missing a lecture for the third year students was therefore less costly in terms of loss in performance in comparison with second year students.

Other separate studies in the US have offered further insights into the relationship between attendance and performance. Brocato (1989) presented the results of a correlation study between attendance and performance of over 400 students studying introductory and intermediate level courses in economics at the University of North Texas between 1983-1987. Attendance was found to be important overall but especially so for students taking the introductory level course. Wiley (1992) used five variables (number of absences, academic classification, age, grade point average and major) to try and predict performance of 186 students studying an introduction to business course. Her results suggested that 57% of the variation in performance was accounted for by levels of attendance. She also found that attending was more important for students in the sample who were taking business studies, rather than some other subject, as their major. The study by Riggs & Blanco (1994) looked at the importance of thresholds of attendance to performance. Low rates of non-attendance did not severely impact on performance, but any of the 197 obstetrics and gynecology students in their study who missed 30% or more of lectures saw a sharp reduction in performance. Thresholds in performance (i.e. grades) were found to be strongly related to attendance in a study of 346 economics students (Durden & Ellis, 1995) who found that grades, rather than actual percentage marks, were statistically more significant in relation to lecture attendance. Launius (1997), on the other hand, found that the performance of 378 students, who had taken an introduction to psychology course during 1990-1995, in mid-semester tests, assignments and final examinations, were affected significantly, but differently, by attendance.

What these various studies reveal is that attendance is found to be associated with performance irrespective of the context in which the students study and the control variables used. The importance of attending may be different for different types of students on different courses, level may also be a differentiator as might thresholds of attendance and type of assessment, but in each of these studies the specific importance of attendance to performance was demonstrated at least statistically.

Method:

The method adopted for this research is similar to these US studies in that data was collected on attendance and performance, with other data included in the regressions as control variables.

The database was compiled as follows. A register of attendance was taken at lectures on a first year undergraduate module at a business school offered in either semester 1 or 2 during the academic year 2000/2001. For each of the students, control variable data was then added from records held centrally by the University. Certain student data was missing, but data was obtained for 419 of the 440 students on this module as follows:

VariableDefinition

PERFThe average mark achieved for all the assessed work (%)

TESTThe average mark for the tests and exam (%)

ASSThe mark achieved for the assignment (%)

LECTThe number of lectures attended

Control Variables

AGEThe age of the student

ABILThe A`level points (or equivalent) on entry to the University

TYPEWhether registered as an HND (0) or degree (1) student

GENDMale (0) and female (1)

ACCOMIf live in University accommodation (0) or elsewhere (1)

MODEIf mode of study part-time (0), or full-time (1)

Notes:

1. (ABIL) Of the 419 students in the database, 132 had either a combination of A`levels and a GNVQ level 3 or only a GNVQ qualification on entry. The current system adopted by UCAS was used in order to estimate the A`level equivalent points for the GNVQ qualification. A Pass was therefore estimated to be equivalent to 6 A`level points, a Merit to 12 and a Distinction to 18.
2. (TEST) (ASS) The marks for assessed work are weighted according to how the module is assessed. The figures for TEST relate to the average mark for the four small tests and end of semester examination that the students take during the module and are worth 60% of the module marks. Tests were not taken in lectures but at another time outside the lecture. ASS is the other piece of work and is an assignment worth 40% of the overall marks. Thus, ASS + TEST = PERF
3. (LECT) As explained in the text, LECT was based on 10 lectures over weeks 3 to 12 of the semester. We considered including tutorials in the database as a separate variable but it proved impossible to keep accurate records as students continued to attend tutorials other than those they were timetabled for.
4. (TYPE) HND and degree students are taught together on this and certain other first year modules at this business school. This offered the opportunity to distinguish between classification of student as Wiley (1992) mentioned earlier did. To qualify to enter as a degree student at this University requires at least two A`levels or equivalent; to study for an HND one A`level or equivalent.

5(ACCOM) Defined as students who live in University accommodation (halls of residence and student houses) and those who live elsewhere either at home or in private accommodation. It could be argued that students in College accommodation are more likely to settle in well to College life and academic study and this may contribute to their performance. There is some evidence that first year students in University accommodation attend more (Wyatt, 1992).

The data collected only included those students who completed the entire module from week 3 onwards. This was done because students at this University are allowed three weeks before they need to confirm which modules they wish to do, and it is known that about 10% attend a number of different lectures and then opt to change between modules in this way. Weeks 1 & 2 are therefore considered a settling in period, particularly for first year students, and are not therefore a true representation of attendance. We also did not include those who attended from week 3 but did not complete the module. This may have introduced a small `survivor bias` as suggested by Chan et al (1997) earlier[8], but to assume leaving the module is simply a reflection of academic failure as they did is debatable. We know from the work of Yorke et al (1997) that students who dropout of University do not necessarily do so because of academic failure, and we might infer the same is true of students who decide to withdraw from a module. To include them in the database might have introduced a bias since their decision to leave the module, and thus their attendance, might have had nothing to do with their performance.