Mobile Edition
Print
Get the Mag
Weekly UpdatesTable 2 summarizes the maximum diversity possible for each measure as a function of team size. It can be seen herein that the maximum does tend to increase as team size increases except for gender diversity, which stays at or close to a maximum of .50. These limitations are worth noting because they affect the range of possibilities for each measure for teams of varying sizes. To control for this effect, the study compares self-formed and randomly formed teams of the same size.
DIVERSITY WITHIN SECTIONS
The diversity within a section is an important factor for determining the level of diversity possible for any team. Ideally, section diversity should neither be excessive nor absent. For example, a section would not be useful for testing this study's hypotheses for ancestry if every student in the section were of a different ancestral background, since there would be no way for students to align in a homogeneous way by ancestry. Such excessive diversity is not problematic for gender and race since there are so few membership categories for these measures and all sections have many more students than the number of categories (range = 17 to 93 students). Section diversity that approached zero would likewise be problematic. Such a section, comprised of students of the same gender, race, or ancestry, would eliminate the possibility of heterogeneity in any team. In sum, gender and culture diversity must exist in the sample studied. However, there cannot be so much diversity that homogeneous assembly is precluded.
In order to check the level of diversity in the eight sections participating in the study, gender, racial and ancestral diversity measures were computed separately for each section as a whole. In the cases of gender and race, the maximum diversity possible for any section is comparable to that for the smaller teams embedded in them. The maximum is .50 and .80 for gender diversity and racial diversity, respectively, varying slightly as a function of the divisibility of section size by the number of categories possible. For ancestry, the maximum varies more directly as a function of section size since there are many potential ancestral categories. For the smallest section, the maximum possible would be 17 students from 17 different ancestries, resulting in a maximum diversity of .941. For the largest section, the maximum possible would be 93 students from 93 different ancestries for a maximum diversity of .989. Of course, in the latter case, it is unlikely that students would spread out over so many categories (nor would such a situation be optimal for testing the study's hypotheses). The minimum possible is zero for each diversity measure, which would occur if every student in the section were the same gender, race, or ancestry.
Table 3 shows the diversity of each section on the three measures. While diversity of the sections does vary, it does not appear that excessive diversity or lack of diversity are important problems in any section except for section 2, where racial diversity is relatively low. Table 3 also shows the number of teams developed from each section. The number of teams and average team sizes are, in part, a function of the number of students in a section, with the average team size ranging from 3.14 to 5.25. Moreover, all sections are of sufficient size to generate useful levels of diversity. Even section 7, which is relatively small, has levels of ancestral and gender diversity comparable to those present in the other sections. Its racial diversity also appears adequate in spite of its smaller size.
RESULTS
The diversity of the self-formed and randomly formed teams on the measures of gender, racial, and ancestral diversity is shown for each section in Table 4. From this table, it can be seen that the self-formed teams are typified by less diversity than the randomly formed teams on all three measures. In every instance, the self-formed teams had lower levels of diversity, although in the case of race for sections 2, 4, and 7, the diversity of the two types of teams are especially close, as is ancestral diversity for section 7. As shown earlier, these sections also had the lowest levels of racial diversity in the study. The lower level of racial diversity present in these sections may, in part, account for the closeness of the two types of teams within these sections on this measure.
To statistically test the study's three hypotheses in a more formal manner, a comparison of the two types of teams was made across the 80 pairs of teams for each of the three diversity measures using the Wilcoxon Matched-Pairs Signed-Ranks Test. This test was used because it makes less stringent assumptions about the distributions of scores than is common in comparable parametric tests such as the paired t-test, including less concern over heterogeneity of variance in the dependent samples (Sheskin, 1997).
Like other nonparametric tests that drop ties from the analysis, the Wilcoxin Test is also more capable of handling in a valid manner the less than continuous nature of the diversity measures obtained under each condition. By dropping ties from the analysis, the power of the test is limited by the resulting smaller sample size, but this does not affect any of the conclusions to be made about the sample (Bradley, 1968). This test is also appropriate for testing two related samples, such as those considered in this study, which are related in the sense that the students used to develop the teams are identical in both study conditions (i.e., self-formed and randomly formed). The Wilcoxon Test is also somewhat more powerful than its counterpart, the Sign Test, since it considers the magnitude of differences and not just the presence of a difference.
In Table 5, statistics comparing the teams (n = 80) on the three diversity measures are shown. Significant differences occurred in the diversity of the two types of teams with self-formed teams showing significantly less diversity for gender (Z = 3.81, p [less than or equal to] .0001), race (Z = 4.500, p [less than or equal to] .0001) and ancestry (Z = 5.13, p [less than or equal to] .0001). The number of tied scores between pairs was 28, 14, and 25 for the measures of gender, racial, and ancestral diversity, respectively.
In addition, the difference in the percentage of self-formed and randomly formed teams having no diversity on the measures was assessed statistically using the McNemar Test for proportion comparison between matched groups. For self-formed teams, the percentage of totally homogeneous groups is 36.3%, 30%, and 21.3% for gender, racial, and ancestral diversity, respectively. For randomly formed teams, the percentages are 8.8%, 2.5%, and 0% for gender, racial, and ancestral diversity, respectively. All three differences are significant (p [less than or equal to] .001), indicating that more self-formed teams are completely homogeneous than are randomly formed teams. This result also supports hypotheses one, two, and three since the high incidence of heterogeneity in the randomly formed teams clearly demonstrates that more diversity could have occurred in the self-formed teams if students had wanted to align in more diverse groupings.
DISCUSSION
The present research has examined the formation of strategic decision-making groups by students in a capstone business strategy course. The subjects in our study were seeking task-based outcomes (grades) from their teams. Therefore, one would expect them to seek a group that would maximize their chances of attaining a high grade.
College students frequently hear and read that diversity can be advantageous for work group productivity. In addition, business school students should be less prone to inappropriate discrimination, or at least more attuned to behaving in a politically correct manner (Brief et al., 1995). Yet in spite of evidence that work groups with gender and culture diversity have the potential to generate a greater variety of ideas and to improve group decision making, especially over time (Watson et al., 1993), students in this study sought to create teams that were highly homogeneous in gender, race, and ancestry.
This behavior is in keeping with the predictions of the similarity-attraction and self-categorization theories. However, a competing argument can be made that the costs of diversity, such as conflict, communication problems, and high turnover rates (O'Reilly et al., 1989), are also known to students, who therefore may have grouped together homogeneously in order to avoid the unproductive concomitants of heterogeneity. From this study, we can conclude only that, when given the opportunity to form diverse groups for decision making, students are disinclined to do so.
STUDY LIMITATIONS AND IMPLICATIONS FOR FUTURE RESEARCH
Since this study used a convenience sample, the generalizability of results is constrained. More research needs to be done that replicates and extends our findings. Though there is replication within this study through the use of multiple sections over time and with different instructors, replication in other settings is an essential next step. For example, given that students may behave differently from experienced workers, it would be beneficial to conduct field research or even non-student laboratory studies.
Another potential problem is the validity of our racial and ancestral diversity measures. Questions about race and ancestry face obstacles from difficulties in definitions. While we used the categories compiled by the U.S. Census Bureau, there were instances in which students checked more than one category for ancestry. In those cases, we had to make subjective judgements about where to place the responses. We chose to include a "mixed" category to accommodate these answers. As previously reported, there were just three such responses. Also, two individuals reported their ancestry in ways that we felt were amenable to being joined into one category. One student reported West Indies and another Jamaica as their ancestral countries and we recorded both under Caribbean.
FEED