The impact of group diversity on class performance: evidence from college classrooms

Abstract

We combine class performance data from an undergraduate management course with students’ personal records to examine how group diversity affects group work performance and individual learning. Students are exogenously assigned to groups. We find that, on average, male-dominant groups performed worse in their group work and learned less (based on their grades in individually taken exams). This gender effect is highly significant in individual learning outcomes providing evidence that gender diversity is influential in the level and nature of knowledge transfers within groups. The results are robust to controlling for the team governance form, a unique feature in our study. Finally, racial diversity had no significant effect on group or individual performances.

Keywords:

• teams
• diversity
• productivity
• learning
• collaborative skills
• educational economics

JEL classifications:

• D2: Production and Organizations
• I2: Education

Acknowledgements

We thank Bill Bottom, Stuart Bunderson, Bart Hamilton, Caroline Hoxby, Marc Law, Jackson Nickerson, Ansgar Richter, Masaru Sasaki, Jacob Vigdor, and seminar participants at the Olin School of Business and the Department of Economics at Washington University in St. Louis, the National Bureau of Economic Research, Higher Education Meetings, the Midwest Economics Association Conference, and the Japanese Economic Association semi-annual meetings for helpful comments and suggestions. Part of this research was completed while the authors were at the Washington University in St. Louis. Any errors are the responsibility of the authors.

Notes

1. Exceptions include Han and Li (2009) who focus on college residential peers whose assignment is known to be random and Hamilton, Nickerson, and Owan (2003, 2012) who use detailed data on worker productivity and demographic information from a garment factory.

2. Past studies have shown that the social context is important in analyzing team performance (Bandiera, Barankay, and Rasul 2010; Bernhard, Fehr, and Fischbacker 2006; Goette, Huffman, and Stephan 2006).

3. In addition, previous research in social psychology literature mostly concluded that gender heterogeneity leads to a higher level of conflicts (Alagna, Reddy, and Collins 1982; Pelled 1996), lower performance in some cognitive team activities (Clement and Schiereck 1973), and less creative work (Hoffman, Harburg, and Maier 1962; Kent and McGrath 1969).

4. Although two different instructors taught the course, the same instruction format was used and the format was kept constant over time.

5. As we describe later, four- or five-member teams were initially formed. When students drop the course after the first week, teams may be left with three or two members. Attrition is unlikely to be associated with the expected performance of teams, however, because most attrition took place very early in the semester before students started working on their group projects.

6. The SAT is the most widely used college admission test in the US. For those students who took the ACT, another standardized test for high school achievement and college admission, instead of the SAT, the ACT scores were converted into SAT scores according to the Comparison Table provided by College Educational Board. The comparison table was constructed based on 103,525 test takers who took both tests between October 1994 and December 1996. Equivalent SAT and ACT scores are those with the same percentile ranks for a common group of test takers.

7. The primary divisions of the students in our sample pool are the schools of arts and sciences, fine arts, architecture, business, and engineering.

8. The discussion sections are not designed to be lecture reviews. TAs guide discussions of business cases or run simulation games that are related to lectures to enhance student understanding of class concepts in a small class setting.

9. For discrete measures of group characteristics such as gender and racial compositions, we also constructed Pearson chi-squared test statistic. We did not find any evidence that group characteristics are biased in light of random assignment except for gender.

10. It should be noted that there are some differences in class characteristics. First-year business students as well as other interested freshmen take the course in the Fall of their first year. Other students, including those with a minor in business, those who are transferring to the business school, and other interested students generally take the course in the Spring. Therefore, age distribution is quite different between the Fall and the Spring classes.

11. Our empirical results are robust and very similar under alternative measures of ethnic diversity such as the number of different ethnic backgrounds within a group.

12. We also estimated models with dummy variables controlling for all-male and all-female groups as well as groups with only one female and one male member instead of using our gender dominance variables. Results are similar to the reported ones, indicating that all-male groups and groups with only one female member perform worse in their group projects.

13. One plausible explanation that reconciles these seemingly inconsistent observations is that group work and individually taken exams differ in their degree to which self-discipline affects performance. Using class and test performance data of eight-grade students at a magnet school, Duckworth and Seligman (2006) find that girls are more self-disciplined than boys and this advantage helps girls to earn higher GPAs (Grade Point Averages) but not so much as to increase their success in achievement or aptitude tests. Similar explanation may apply to college we study. Group assignment in MGT100 requires many days and hours of hard work and collaboration to produce high grades, which require higher degree of self-discipline, whereas attending classes and learning problem-solving skills through homework may be sufficient to get high exam scores.

14. A cursory look at the differences among genders in SAT verbal and SAT math scores in our sample indicated that female and male students performed similarly in the verbal section of the SAT; however, male students’ math scores were statistically significantly higher than the female students’ math scores.

15. Although not reported in Table 3, we also estimated model in column (2) using only Fall 2004 class, which is made entirely of first-year college students. Supporting our maturity hypothesis, we find that the estimated coefficient of malegroup variable is statistically significant at 5% level and larger (−13.81) than the same estimated coefficient (−7.33) in the full sample.

16. The two midterm exams each account for 20% of the course grade and the final exam accounts for 25%. IndScore is calculated based on these weights, that is, IndScore = (20*midterm1 + 20*midterm2 + 25*final)/65.

17. We also estimated an ordered Probit model to take into account the possible nonlinearity between performance measures and group and individual characteristics. Efforts to raise grades from 60% to 70% are likely to be different than efforts to raise grades from 80% to 90%. However, results of ordered Probit estimation are qualitatively very similar to the OLS estimation and, thus, are not reported.

18. We also measured age heterogeneity (and knowledge diversity) within groups using the ratio of maximum and minimum age among group members. This measure did not perform as well in our models, but produced similar results.