‘D’ is for dilly-dally?

Abstract

Evidence from online assignments in an intermediate microeconomics course suggests that nonprocrastinators (both early-starters and front-loaders) score higher than their dilly-dallying counterparts. Students who are busier in school tend to start their assignments earlier.

Acknowledgements

We acknowledge the support of Lyssa Enzmann at Aplia for providing the data that has enabled the creation of our early-starter and front-loader variables and Sanjib Sarker for research assistance. We thank participants in Utah State University's Department of Economics seminar series for useful comments on an earlier draft of this article. We also thank the students who participated in this study.

Notes

¹ The closest studies to ours concerning the factors influencing student performance in economics are Johnson et al. (2002) and Krohn and O'Connor (2005), but these deal with measures of effort rather than delay.

² What a professor perceives as dilly-dallying may instead be the optimal outcome of the student's time-allocation problem. In this case, the cost incurred is that of optimal rather than abject dilly-dallying. To distinguish empirically between the two would require an underlying general equilibrium model of student choice.

³ The assignments were advanced 'workouts' in Bergstrom and Varian (2003). The course text was Varian (2003). Twenty-three students completed the assignments. The course began with 25 students, thus it is unlikely that our data suffers from missing data, as noted in Becker and Powers (2001).

⁴ HRSWORK is self-reported and thus may be subject to the ‘Lake Wobegon Effect’ (Maxwell and Lopus, 1994). However, the typical USU student is a member of the LDS Church, which encourages its members to begin their married lives and careers earlier than the national average. We also compared self-reported GPA and CREDITS with official values from transcripts. Mean values for official and self-reported GPAs were 3.34 and 3.37, respectively, not statistically significantly different at the 95% level. The official and self-reported means for CREDITS were 14.25 and 12.05, respectively. This under-reporting of credit hours was significant at the 95% level.

Table 1. Variable definitions

⁵ This measure cannot account for the relative difficulty of questions and thus may overstate the loading behaviour of some students (for those who back-loaded their effort by having started the more difficult questions first).

⁶ We tested the standard model for heteroskedasticity and within-panel (AR1) autocorrelation using feasible GLS (Greene, 2003). Results correcting for these possible error structures were qualitatively similar to those without the corrections, which are reported below. Alternative regressions were considered with a normalized score to control for difficulty across assignments, but results were again qualitatively similar. Panels are unbalanced due to a few missing values for the dependent variable, SCORE. Missing values occurred when students chose not to attempt some of the graded assignments.

⁷ For the standard model, the LM test rejects the pooled OLS model in favour of random effects, and the Hausman χ² test rejects FE in favour of RE. For the IV model, the Hausman χ² test similarly rejects FE in favour of RE.

⁸ Our results also suggest that, all else equal, students who completed practice assignments and had higher GPAs similarly performed better. Students who worked longer hours outside of school performed worse. The large coefficient for GPA could reflect the fact that the Aplia assignments are considered ‘advanced’ by Bergstrom and Varian (2003). Or, GPA could be proxying for other unmeasured determinants of overall student capability.

⁹ To test whether procrastination has a negative effect on exam scores, we also ran a simple OLS regression of total exam points on GPA, GENDER, CREDITS, HRSWORK and the averages of SKEW and START across the nine assignments. GPA had a strong positive effect, as did CREDITS. HRSWORK had a negative effect. However, averaged SKEW and START had no statistical effect. Thus, procrastination on homework assignments does not necessarily translate into poorer performance on exams.

¹⁰ In the case of front-/back-loaders, we find no evidence that an additional credit hour or an additional hour worked induces more back-loading. However, students did back-load their effort most during the middle of the semester. This could reflect conventional wisdom that students start the semester full of enthusiasm and finish in a state of trauma, which reduces their penchant for back-loading. Results are available from the authors.