Do Monetary Incentives Matter in Classroom Experiments? Effects on Course Performance

Abstract

Using 641 principles of economics students across four universities, the authors examine whether providing monetary incentives in a prisoner's dilemma game enhances student learning as measured by a set of common exam questions. Subjects either play a two-player prisoner's dilemma game for real money, play the same game with no money at stake (i.e., play a hypothetical version), or are in a control group where no game is played. The authors find strong evidence that students who played the classroom game for real money earned higher test scores than students who played the hypothetical game or where no game was played. Their findings challenge the conventional wisdom that monetary incentives are unnecessary in classroom experiments.

Keywords:

• classroom experiments
• course performance
• monetary incentives

JEL codes:

• A20
• C90

ACKNOWLEDGEMENTS

The authors are grateful for helpful comments at the 2013 AEA Conference on Teaching and Research on Teaching, the 2013 Southern Economics Association Conference, the 2014 ASSA Conference in Philadelphia, the 2014 Appalachian Research in Business Symposium, and from seminar participants at Susquehanna University. The authors also thank two anonymous reviewers for their helpful feedback.

Notes

¹The concept of individual response to incentives, both positive and negative, is a cornerstone of many widely adopted principles of economics textbooks. Mankiw (2014) highlighted the power of incentives as one of his core “Ten Principles of Economics” and built upon this principle throughout the textbook.

²In the game, we called the choices “share” or “hoard” as we felt those terms were easier for students to understand.

³The actual test questions used and instructions for the game are available from the authors upon request.

⁴While the specific method of grading the questions varied across professors, that should not affect the variation in student scores for any given professor.

⁵We also ran non-parametric Wilcoxon rank-sum tests, which showed results that were virtually identical to the t-tests. Results are available from the authors upon request.

⁶The share of the total possible points of the game theory assessment questions on the exam varied across professors but was approximately 15 percent for all but the Kenyon students, who took a separate quiz solely on the game theory assessment questions. For Kenyon students, when we did comparisons of the score on the game theory questions to other exam questions, we used the scores students earned on the previous exam.

⁷We also estimated the same four regressions shown in table 2 using the students’ score on the nongame theory portion of the exam as the dependent variable. We found no statistically significant effect of being in the real money group versus other groups in three of the four models, but found a marginal effect (p <.10) in the remaining model. This provides modest evidence that while the game for real money stakes improved learning on the game theory questions, it appears that there were not significant spillover effects to other topics within the course. Results are available from the author upon request.

⁸The t-values for tests to examine differences between the hypothetical and real groups for the three models shown were 2.31, 2.30, and 1.84, meaning that in models 1 and 2 the difference is statistically significant at the 5-percent level, while in model 3 the difference is statistically significant at the 10-percent level.