Evidence from ANS2012 on how student attendance enhances academic performance

In a paper published in the Economics of Education Review in 2012, Warwick Economics professors Wiji Arulampalam, Robin Naylor and Jeremy Smith (ANS2012) examined whether students who missed classes were likely to achieve high marks in their university courses than students with better records of attendance. They found evidence of adverse effects of absence on performance.

Methods and findings of the paper

The dataset is based on 1332 observations of second-year students taking three core modules as part of their Economics degree course at a particular university in the UK. The measure of academic performance is the student’s end-of-year mark in each of their 3 courses (micro, macro and econometrics). The data include information on each student’s record of attendance in each of their weekly seminar meetings in each of their courses. The authors find that, on average, a student’s mark in a module is higher if they miss fewer seminar meetings in that module. This seems to be driven by absence being particularly harmful for students with greater potential to perform well, suggesting that better-performing students are also the ones who tend to gain most from attending class.

But is the apparently positive effect of attending class really a causal effect? We know that “correlation does not necessarily imply causation.” Are the results not simply a consequence of more able and more hard-working students being (a) more likely to achieve high marks and (b) more likely to attend class – without class attendance itself having a causal effect on marks. In other words, does the observed correlation merely arise from the fact that attending class is a choice variable with better-performing students selecting to attend class – thereby generating a ‘selection bias’ in the estimated effect of attendance on performance?

The authors argue that their methodology enables them to overcome such potential biases and hence that they have identified a causal effect of attendance on performance. One of their justifications for this is based on the fact that the observed students do not themselves sign up to seminar groups for their three modules: instead, students are allocated randomly to groups and to time slots on the weekly timetable. The authors argue that this generates exogenous variation in a student’s attendance. For example, if you have a class at 9am and your bus is running late, then the fact that you miss the class is not a choice variable for you. Exogenous variation in class attendance overcomes the selection bias problem in the same way that randomised control trials (the gold standard for identification of causality) attempt to ensure that the treatment group and the control group do not vary systematically.

Conclusion: Work hard, don't skip class, get better grades!