Demographic groups respond differently to incentives depending on the stakes of the task they are undertaking.
This is the conclusion of new research by Analia Schlosser, Zvika Neeman and Yigal Attali published in the October 2019 issue of The Economic Journal, which looks at how performance in high and low-stakes tests differ across gender, race, and ethnicity.
Interest in these questions stems from attempts to explain gender, racial, and ethnic imbalances in human capital accumulation and labour market performance. Increased use of aptitude tests for college admissions and job screening, as well as the growing use of standardised tests for the assessment of students’ learning, have also contributed to this surging interest.
In their study, Schlosser and colleagues examine whether individuals respond differently to incentives by analysing their performance in “high” and “low” stakes situations. The high stakes situation is the Graduate Record Examination General Test (GRE) that students take as part of the admission process to graduate studies in the US and most English-speaking countries. The low stakes situation is a voluntary experimental section of the GRE that examinees were invited to take part in immediately after finishing the real GRE examination.
The authors compare the change in performance between a high stakes setting that has important consequences and a task that has almost zero stakes by gender, race, and ethnicity.
The results show that males exhibit a larger decrease in performance between the high and low stakes test than females and that white people exhibit a larger decrease in performance between the high and low stakes test compared to Asians, black people and Hispanics.
The findings are partially explained by lower levels of effort exerted by males and whites in the low stakes situation compared to women and minorities, respectively. The authors do not find evidence supporting alternative explanations such as test anxiety or stereotype threat.
The findings show that differences in the perceived importance of the test can significantly affect the ranking of individuals by performance and they have two main implications:
- Stakes have to be taken into account when analysing performance gaps between groups
- Some groups (males, whites) are mostly driven by incentives while other groups exert high effort even if stakes are low or “nearly zero”.
These two implications are highly relevant for policy. They stress that any analysis of gender or race test score gaps, or impact evaluations of the effectiveness of educational interventions by gender or race, should consider the stakes of the test involved in order to interpret the results.
In addition, the results highlight the fact that university or job admission policies that use standardised aptitude tests should consider that such tests measure only performance under a high stakes setup and are less informative about individuals’ performance in low stakes or zero stakes situations, which may be as important at the university or job.
The results also have important implications for personnel and incentive policies as they suggest that differences in productivity between workers could vary according to the incentive scheme attached to the job. Often, job candidates are evaluated using high stake tests. However, most jobs require excellence in tasks that are not directly attached to high-powered incentive schemes. Therefore, consideration of performance in high stake tests should not come at the expense of other indicators of ability that reflect good performance in low stake situations.
Differential Performance in High vs. Low Stakes Tests: Evidence from the GRE test by Analia Schlosser, Zvika Neeman and Yigal Attali is published in the October 2019 issue of The Economic Journal
