Examining grade loss aversion – the importance of reference points

ABSTRACTGoal setting and grade loss aversion have individually been found to be effective motivators for student performance. We combine these two strategies to investigate, in a university setting, the effectiveness of grade endowment within a goal-setting framework. Our participants set grade goals which some must earn the traditional way, thus framed as ‘gains’, while others are awarded their target grade upfront, thus framed as ‘losses’. We find a positive impact of goal setting on student performance. We also find that the effect of grade goals framed as ‘losses’ is numerically, but not statistically, greater than those framed as ‘gains’. AcknowledgmentsThe study was approved by the Institutional Review Boards at both participating institutions (IRB #2019-20-0009 and #965). All human subjects provided informed consent.Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 We note an unconventional specification of the interaction term in Smith et al. (Citation2019) to detect gender effect. ‘Male’ is included as a dummy but the interaction term is specified as ‘female × counting down’, rendering the identification of a gender effect confusing, if not complicated.2 Logit regression is used for dependent variables that are indicators and not continuous. We also employ robust standard errors in each regression.3 It is customary, in student performance analysis, to include grade point average (GPA) as an explanatory variable. GPA was not consistently available for all participants since many were either freshmen or had transferred from a two-year college and had no institutional GPA yet. Results are expected to be robust to controlling for this variable since GPA is not expected to have a time-varying effect on performance.4 Data on age is missing for one control group section and so this variable does not appear in the results section. The maximum age observed, 29 years old, is an outlier where the next highest age is 25. Deleting this outlier results in the data passing this balancing test and no substantive change in the results presented in the next section. These results are available upon request.5 Estimations employing clustered and bootstrapped standard errors were also obtained. No notable difference in inference was observed. These results are available upon request.Additional informationFundingThis research received no specific grant from any funding agency in the public, commercial, or non-profit sectors.