Exogenous and Endogenous Factors Influencing Students' Performance in Undergraduate Accounting Modules

Abstract

The objective of this study is to explore cross-sectionally the potential determinants of academic performance in undergraduate accounting modules for non-accounting degree students. The econometric model used in this study employs both student-oriented endogenous factors (e.g. age, gender) as well as student-exogenous factors (e.g. quality of teaching, examination structure). The univariate and multi-variate analyses indicate that both factors play a significant role in determining students' performance. The findings based on the implications of endogenous factors could provide some recommendations about university student admissions and budgeting policies, whilst those based on the implications of exogenous factors could help review students' assessment and academic staff recruitment policies, as well as the quality of teaching.

Keywords:

• Performance
• exogenous factors
• factor analysis
• non-specialist
• undergraduate accounting

Notes

Bartlett et al. (1993) contend that the advantage of having prior accounting knowledge can only be seen in the early stages of undergraduate study. The studies that focus on the impact of previous accounting knowledge on the first year accounting marks are Bartlett et al. (1993), Doran et al. (1991), Gul and Fong (1993), Naser and Peel (1998) and Rankin et al. (2003). Jackling and Anderson (1998) and Rohde and Kavanagh (1996) examine the second year accounting performance while Gammie et al. (2003b) and Koh and Koh (1999) examine final year students.

Some studies (e.g. Kember and Gow, 1991) imply that Chinese students tend to adopt surface learning (focus on the learning material) rather than deep learning (focus on the meaning of understanding). If adopting the latter approach brings higher achievements, then the success of Chinese students in business accounting is a paradox. It may be that this distinction is not relevant for Chinese students or they adopt another learning style named as strategic/achieving approach by which students attempt to excel in assessed work (Entwistle, 1987).

The correlation matrix reported in the Appendix is for the variables used in the multi-variate analysis only (restricted model). The correlation matrix including all variables in Table 1 (unrestricted model) is available on request.

Using BA and BC marks separately as alternative dependent variables did not change the quality of results. As background information, the content of curriculum and the nature of assessments of the accounting modules in the school are mainly quantitative. The mode of module assessments is mixed, as it comprises mini-tests, assignments (case studies), homework and a comprehensive final examination.

The findings in Table 5 are based on the restricted model. To obtain this model, (a) 13 exogenous factors in the unrestricted model were converted into three variables using factor analysis; (b) ‘Part-time work, Learning disability, and Personal problems’ factors constituted ‘Inconvenient effects’ variable; (c) ‘Work experience and Academic experience’ factors constituted ‘Experience’ variable. The discussion of the results mainly revolves around the restricted model and, where necessary, we refer to the unrestricted model.

Age may significantly impact performance in the first year only (Lane and Porch, 2002). We thus considered the possibility of an inverted-U shape relation between performance and age. To test for this, we constructed the Squared-age variable, which is the square of Age. The performance is expected to be positively correlated with Age and negatively correlated with Squared-age. This prediction implies that at the lower spectrum performance increases at a decreasing rate as students become more mature and then decreases at the higher spectrum. The results (not reported) showed that the coefficients had the expected signs but Squared-age's coefficient was insignificant. Another set of unreported results showed that using the natural logarithm of student age or using a dummy variable (1, if a student is 21 or over; 0, otherwise) would not change the results as the coefficients on both variables were significantly positive.

To explore this insignificant relation contrasting with the information integration theory, we tested whether the association between performance and study effort is non-linear (Johnson et al., 2002). The estimations showed the absence of a quadratic relation. The presence of externality effect in the form of public good from high achievers to low achievers may lessen individual scores according to ability. This externality could even be negative depending on the degree of competitiveness of the class environment (Cheo, 2003), which may explain why the present study reports an inverse relation between effort and grades.

The students holding top A-level grades in mathematics performed no better than others in accounting. It may be that the impact of increased numeracy skills tends to diminish at advanced levels. One way to control for this non-normality is to add ‘squared-GCSE’ to the model. The results showed that the coefficient of GCSE is significantly positive and that of squared term is negative but the significance level is no less than 10%. This still might explain why obtaining A-level grades in mathematics does not help much in getting relatively high marks in accounting.

Average-accounting (average score of thirteen accounting questions) could be deemed as a comprehensive factor that measures overall student satisfaction in accounting modules. The coefficient estimate of this variable was found to be strongly positive, which emphasises the importance of exogenous factors for students while learning. We also attempted to measure the overall independence of students from exogenous factors by taking the absolute difference between Average-accounting and neutral value of 3 based on the Likert-scale. It was assumed that larger deviations from the neutral value would imply more relevance of exogenous features on learning and thus on academic performance. This variable was incorporated into the model as a proxy for student autonomy. The regression results (not reported) showed that the coefficient estimate of this proxy was statistically insignificant.

We re-ran model (5) in Table 5 for the second and third year students separately to detect any possible distortions in the results. The regression results (available on request) are very similar with a few exceptions. Compared to the full sample results, the variables ‘Age’ and ‘Lecturers and assessment’ turn out to have insignificant coefficients for the second year students. Furthermore, the negative impact of ‘Study effort’ on performance appears to be significant for the second year students only as the underlying coefficient is positive and insignificant for the third year students. The coefficient on ‘Teaching material’ is significant only for the second year students. Finally, partitioning the sample produced insignificant coefficients on ‘Attendance’ for both subgroups.