Mobile Edition
Print
Get the Mag
Weekly UpdatesAll students were provided with a description of the study and asked whether they would agree to participate. No student refused. Eight different sections, taught by three different instructors, participated over a three-year time period. Data were collected every semester from the fall semester of 1995 through the first summer session of 1998. The mean age of students was 24.5 years (range = 20 to 50 years) with 37.7% and 62.3% designating themselves as male and female, respectively.
The racial and ancestral mix of the sample is shown in Table 1, where it can be seen that it is rather diverse on both variables. Data were gathered by self-report. In the case of race, students could choose from a list or write in an answer. In the case of ancestry, students were asked to write in the ancestry with which they most identified themselves. No checklist of possible ancestral identifications was provided since it was felt that an open-ended format would be less likely to foster perceptions of investigator bias for this highly sensitive question. In both cases, students could indicate a multiracial or multi-national background if they so desired.
The issue of multi-racial and multinational identification has become increasingly salient to the manner in which the federal government surveys and counts racial and ethnic populations (Fisher, 1998). However, given the particular nature of our student population, we did not anticipate comparable problems with this issue (i.e., individuals who do not identify with any ethnic category because a single ethnicity falls to fully capture their own self-categorization.) We knew that the vast majority of our students were either recent immigrants or first-generation Americans and were confident that they would endorse few, if any, mixed alignments of either type. In fact, 76 percent of those participating in this study were recent immigrants or first generation Americans. Of these, most were from Russia and China, cultures in which there is currently little intermarriage.
Fortunately, our estimate that unitary alignments would predominate in student endorsements of racial and ancestral backgrounds came to fruition. Had this not been the case, this study would have been more difficult to implement since complex decision rules would have been necessary for determining how to categorize mixed alignments and for deciding when mixed alignments in a given group would be scored as similar or dissimilar. In fact, no student endorsed a mixed racial categorization and only three students listed multiple ancestral identifications with no obvious dominant identification. Thus, we had a preponderance of students identified with one particular ancestral background.
It should be noted that, in the case of ancestry, any alignment listed was used as a category even if it was broader than a specific country (e.g., South American), since it was the student's ancestral identification that was of primary interest. With regard to the three students with a mixed alignment, none ended up in the same group. As a result, all contributed a degree of heterogeneity to their groups and no rules were necessary for determining the level of similarity for two or more members with "mixed" ancestral backgrounds within a single group.
We used the categorizations from the current U.S. census for both ethnicity and race. However, because we did not include separate questions for ethnicity and ancestry, we folded "Hispanic" into the race question. There is precedent for this practice. Scholars and pollsters often use the concepts of race, ethnicity, and nationality interchangeably. It is common in surveys and research instruments to ask respondents to indicate their race by choosing from a list that includes races, ethnicities, and national origins (Betancourt and Lopez, 1993). Interestingly, the new United States census system, which will first be used for Census 2000, will place the ethnic question (e.g., Hispanic, Latino) before that of race in the hope of reducing the number of respondents who choose "other" as their race (Fisher, 1998).
Though the information is not included in Table 1, the survey also queried whether the student and each parent were born in or outside the United States. This information was not used in the study except to confirm our assumption that the majority of our student population is first-generation American or immigrant. As regards to birthplace, 49.2% of the students were born in the United States while 50.8% were born outside the United States. Sixty-eight percent had at least one parent born outside the United States while 32% had both parents born in the United States. In combination, 26% of the sample was comprised of students and parents who were born in the United States. The remaining 74% involved at least one parent or student born outside the United States.
SELF-FORMED TEAM DEVELOPMENT
At the start of the semester, students were told that the course requirements included a computer-based game in which student teams would manage business units with the goal of maximizing competitive standing, based primarily on the firm's return on assets. About one third of the way through the semester, students were told to assemble into teams for the simulation. The only direction from the instructors was a six-person limit on team size. Absent students were permitted to join any team they wanted and that wanted them when they next came to class as long as this did not violate the maximum team size of six.
Typically, there is no turnover in this capstone course besides those students who drop or add the course within the first two (or sometimes three) weeks of the semester. Therefore, groups were intentionally assembled a few weeks into the semester to assure that the rosters would remain stable. All students on all teams completed the course.
Across the eight sections of the course, a total of 80 teams were formed ranging in size from one two-member team to nine six-member teams, with an average team size of 4.48 members. Eighty-eight percent of the teams were three-, four- or five-member teams with five-member teams being the most common size at 43.8 percent of the total. These teams are referred to as "self-formed" teams in this study.
RANDOMLY FORMED TEAM DEVELOPMENT
In order to determine the level of group diversity that would occur if students had been organized into teams of exactly the same size in a random fashion, 80 parallel, "phantom" teams were developed from the roster of students within each section. The only difference between these teams and those developed through self-selection lies in the method of group formation and not in the underlying distributions of gender, race, or ancestry since both sets of teams are formed from the same students.
The term "phantom" applies to these teams because they existed as teams only "on paper." Their purpose was to provide a baseline measure of the level of diversity possible in each section's teams, absent any concerted effort to create diversity. These phantom teams are referred to as "randomly formed teams" in this study.
GROUP DIVERSITY MEASURES
Demographic information was collected from team members though a self-report inventory. This inventory was administered to students on a day near the end of the semester in which class attendance was expected to be high due to class requirements. Absent students were tracked through follow-up procedures on another class day in order to minimize the amount of missing data in the study, with all students eventually completing the inventory. To minimize any social impact that teams might have on inventory completion, students completed the inventory individually and not while they were working in their teams.
Group diversity measures for each team were computed for gender, race, and ancestry using BIau's (1977) index of heterogeneity as the basis for each measure. This index has been recommended by a number of investigators as an important measure of group diversity (Lau and Murnigham, 1998; Jackson, et at., 1991).
Blau's index can range from 0 to a high of 1. It is defined as follows:
Heterogeneity = ( 1 - [sigma] [[p.sup.2].sub.i])
where [P.sub.i] is the proportion of a given nominal-scale group i represented within the larger group of interest. The low of 0 occurs when all members of a group are identical on a particular measure. For example, a team comprised of four females would have a score of 0 for gender diversity. Likewise, a team comprised of six Asians, all of Chinese ancestry, would have a score of 0 for racial diversity and a score of 0 for ancestral diversity.
Blau's index does not necessarily reach a maximum of 1 for all categorical variables. It approaches 1 as the size of a team increases and as the subsets within the team become evenly distributed. As its value increases, the dispersion of group members across the potential values for a given categorical variable increases, the dispersion of group members across the potential values for a given categorical variable increases, leading to a greater probability of contact between diverse members of the group. For such narrow categorizations as gender and race, the value of 1 cannot be reached even with large groups. For these, the maximum diversity possible is limited not only by group size but also by the limited number of categories possible. For example, a two-person team comprised of one male and one female is at maximum dispersion and would have a gender diversity of .50 (i.e., 1 -- ([.50.sup.2] + [.50.sup.2])). Indeed, the maximum gender diversity possible for a group of any size is .50.
Racial and ancestral diversity for a two-person team also reach a maximum of .50 when the team is comprised of two different races or two different ancestries. Gender diversity for a three-person team is at its maximum when the team is comprised of two of one gender and one of the other. So, a three-person team comprised of two males and one female would have a gender diversity of .444 (i.e., 1 - ([.667.sup.2] + [.333.sup.2])). However, such a team could have three different races or three different ancestries represented and as a result reach a maximum diversity for these measures of .667 (i.e., 1 - ([.333.sup.2] + [.333.sup.2] + [.333.sup.2])). The maximum diversity for race is also limited by the limited categories possible. Racial diversity can only reach a maximum of .80 for any group size. Only in the case of ancestry, where many categories exist, can a maximum diversity of 1.0 be achieved.
FEED