Modeling Accentuation Effects:Enrolling in a Diversity Course and the Importance of Social Action Engagement
Using structural equation modeling, our analysis of survey data collected from two diversity courses and a management course suggests that enrollment in diversity courses increases students' positive interactions with diverse peers and "accentuates" the importance students place on engaging in social action.
Institutions of higher education have undertaken and have continued to create numerous programs and initiatives to increase students' engagement with social diversity. By creating contexts in which students encounter others from different backgrounds, interact across those differences, and learn about diverse peoples and perspectives, these programs help students learn the knowledge and skills necessary to function effectively in our increasingly diverse and complex society. The Association of American Colleges and Universities [AACU] (1995a) articulated this message in their American Commitment's Initiative, which states that today's college students "must learn, in every part of their educational experience, to live creatively with the multiplicity, ambiguity, and irreducible differences that are the defining conditions of the contemporary world" (p. xxii). In light of the events of September 11 and the concomitant increase in racial and ethnic tension around the globe, this statement takes on a broadened meaning and an increased sense of immediacy. [End Page 448]
Although many institutions of higher education have increased the numerical representation of racial/ethnic groups on their campuses over the last several decades, American society has actually increased its racial and economic segregation (Orfield, Bachmeier, James, & Elite, 1997). Consequently, students frequently enter relatively diverse college environments from highly segregated high schools. Evidence suggests, however, that engagement with diversity on college campuses significantly increases the chances that a person will choose to live and work in a diverse environment, thereby minimizing the resiliency of the segregation trend (Gurin, 1999).
One of the important locations of students' engagement with diversity is the college classroom. The classroom environment offers potential opportunities for students to learn to negotiate and communicate across difference while overcoming the inherent challenges in working with diverse groups. Mounting evidence suggests that incorporating diversity into the curriculum, through both content and structured interactions among diverse peers, makes students more likely to increase, among other things, the complexity of their thinking and their willingness to engage current social problems (see Gurin, Dey, Hurtado, & Gurin, 2002; Smith & Associates, 1997).
Despite a growing number of national reports and committees focusing on issues of civic renewal, minimal attention is paid to the role that higher education can and should play in promoting the development of civic values and responsibility (Colby, Ehrlich, Beaumont, Rosner, & Stephens, 2000). Boyer (1996) asserts that excessive individualism, faulty reward systems, and a de-emphasis on service in many research universities foster a climate of disengagement and move higher education away from its larger purpose: "to participate in the building of a more just society and to make the nation more civil and secure" (p. 13). Further, a climate of disengagement leaves students unprepared for their roles as democratic decision makers in an increasingly diverse society with complex social problems.
In order to respond to these growing concerns about the decline in civic values among college students, many colleges are making renewed efforts to impress upon their student bodies the importance of understanding and taking action to address social issues and problems. The AACU (1995a), for instance, is actively exploring how student engagement across difference helps students appreciate multiple perspectives, value human and cultural diversity, and make informed choices. The purpose of this study is to examine how particular courses that are inclusive of diversity in their content and methods of instruction, what we will call "diversity courses," promote the importance students place on [End Page 449] taking personal responsibility for social issues and problems. The extent to which diversity courses increase students' engagement with social issues may serve to offset their current disaffection for civic engagement—a change championed by social theorists such as Putnam (2000) and Barber (1997). Further, increased involvement with social action will likely provide students with the tools needed to navigate a "multicentric democracy"—a democracy that recognizes both the connectivity and distinctness inherent in a community predicated on multiple and simultaneous forms of identity, power, and privilege (Guarasci, Cornwell, & Associates, 1997).
Review of the Literature
The conceptual framework developed for this paper is based on a set of relationships played out over two periods of time, upon entry into a course and at the end of the term. (The specific model, represented by Figure 1, is described in more detail in the next section.) Specifically, we hypothesize that by enrolling in a diversity course, students are likely to report having more "positive" interactions with diverse peers, peers whose race is different from their own. We also contend that both enrollment in a diversity course and increased amounts of positive interactions with diverse peers will increase the level of importance students place on taking action to address social issues and problems, what we call social action engagement. The following review of the literature summarizes the previous research that shaped our understanding of these relationships.
Diversity courses are intended, in part, to prepare students to be effective citizens in a diverse society by encouraging interaction with diverse peers and promoting what we call democratic engagement. Higher education institutions nationwide have made a commitment to teaching cultural pluralism as an important component of the undergraduate curriculum (Olguin & Schmitz, 1997). This commitment is evidenced by reports indicating that 63% of campus mission statements reference diversity as an educational goal (AACU, 1995b). In addition, Levine and Cureton (1992) report that nearly 50% of campuses have ethnic and women's studies programs, and Humphreys (2000) reports that 62% of campuses already have or are in the process of developing a diversity requirement.
For the purposes of this study, diversity courses are those courses that have content and methods of instruction that are inclusive of the diversity found in society. Examples of such courses can be found in [End Page 450] departments such as women's or ethnic studies or under an institution's "diversity" general education requirement. As has been noted by several scholars (e.g., Banks, 2001; McIntosh, 1990), there are numerous ways that courses can include diversity. These different ways can be arranged along a continuum from not at all inclusive to completely inclusive of diversity (in the multicultural education literature, the continuum ranges from monocultural to multicultural). Although applied most frequently to course content, this continuum relates to all elements of a course, such as purposes and methods of instruction (Nelson Laird, 2003). In this study, we are comparing diversity courses that we know to be highly inclusive of diversity in most aspects to a course that is barely inclusive of diversity at all.
There are instances where diversity courses are taught from an additive perspective, one that views the inclusion of women or people of color as an additional, often disruptive, element to what is important to teach (Banks, 2001). Although there is variation, the literature suggests that many diversity courses introduce content and methods of teaching that are intended to expose students to multiple perspectives on issues, to teach them to think more complexly, and to engage them actively in an exploration of forms of oppression within society (Chang, 2002). While participation in courses taught from an additive perspective may not have the same effects as those that are more inclusive of diversity, results from several studies suggest that, taken together, diversity courses have a positive effect on learning, democratic, and multicultural outcomes.
Astin (1993a) and Villalpando (1994), for instance, found that ethnic studies courses, cultural awareness workshops, cross-racial socialization, and discussions of racial issues were associated with widespread beneficial effects on a student's academic and personal development, regardless of the student's race. Furthermore, enrollment in an ethnic or women's studies course was shown to be positively associated with gains in learning outcomes such as complex and sociohistorical thinking (Gurin, 1999), developing critical perspectives (Musil, 1992), foreign language skills (Astin, 1993a, 1993b) and critical thinking (Gurin, 1999; Hurtado, 2001a; Tsui, 1999). Enrollment in these courses was also shown to predict positive changes in democratic outcomes, including promoting racial understanding (Astin, 1993a; Gurin, 1999; Milem, 1994), interpersonal skills (Hurtado, 2001a), and participation in a community action program (Gurin, 1999). In addition, taking a diversity course is related to multicultural outcomes such as reducing prejudice (Chang, 2002), increasing cultural awareness (Astin, 1993b; Gurin, 1999; Hurtado, 2001a), tolerance (Hurtado, 2001a), and awareness of inequality (Lopez, 1993). [End Page 451]
Of those studies identifying the link between diversity courses and democratic outcomes, only a small subset includes measures that examine students' level of involvement with social issues and problems (e.g., Gurin et al., 2002; Hurtado, 2001a). The Gurin et al. (2002) study is particularly convincing, as its findings were consistent across several different national and institutional data sources. Additionally, the large sample sizes and the ability to control for student background characteristics and other confounding variables are strengths of these studies. It is important to note, however, that these studies are rarely based in classrooms; these studies therefore only address the cumulative impact of diversity courses.
Only a limited number of classroom-based studies exist that report results specifically from diversity or diversity-related courses. Of these studies, most reported changes in students' attitudes and beliefs. For example, studies found that diversity courses positively influenced students' racial attitudes (Bakari, 2000; Tran, Young, & Di Lella, 1994), comfort dealing with diversity (Barry, 1996), beliefs about cultural diversity (Moore & Reeves-Kazelskis, 1992; Pedras, White, & Schmidt, 1996; Torok & Aguilar, 2000), empathy (Carrell, 1997), cultural sensitivity (Bakari, 2000; Nel, 1992), and attitudes toward multicultural education (Adler & Confer, 1998; McMahon & Reeves, 1999; Olmedo, 1997).
While the consistency found among the classroom-based studies (almost all report positive effects) is encouraging, the limitations of these studies are noteworthy. Most notable are the lack of comparison groups and the absence of gender and ethnic diversity among participants, which restricts the usefulness of these results beyond the particular course being studied. Furthermore, previous studies mainly focus on education courses and use cross-sectional sampling techniques, thereby lacking a control for the fact that particular students with predispositions toward these outcomes often take diversity courses as a matter of choice and preference. This predisposition to choose courses that are related to an established set of interests, skills, and values may be described as a tendency toward accentuating these characteristics and is inherent in course-taking behavior among students.
Interactions with Diverse Peers
Building on a classic line of research demonstrating the positive effects of intergroup contact (e.g., Allport, 1954; Amir, 1976; Cook, 1984; Pettigrew, 1991), several studies have established how student interactions with diverse peers—White students and students of color [End Page 452] interacting with racial groups other than their own—are connected to educational outcomes important to the mission of higher education (Hurtado, Milen, Clayton-Pedersen, & Allen, 1999). For example, studies have found that interactions with diverse peers positively influence intellectual and social self-confidence (Chang, 1996), openness to diversity and challenge (Pascarella, Edison, Nora, Hagedorn, & Terenzini, 1996), gains in critical thinking (Pascarella, Palmer, Moye, & Pierson, 2001), self-reported gains in problem-solving skills (Hurtado, 2001a), and leadership and cultural knowledge and understanding (Antonio, 1998). In most cases, these results hold after controlling for race or are significant within separate analyses conducted by racial group. As in the current study, several of these works were designed so that meaningful statements could be made about the effects of interactions with diverse peers for students of color as well as for White students.
A subset of these studies has established that interaction with diverse peers predicts democratic outcomes like social action engagement (e.g., Gurin et al., 2002; Hurtado, Engberg, Ponjuan, & Landreman, 2002). In addition, there is a growing body of research linking intergroup relations programs with social action engagement (see Hurtado, 2001b; Zúñiga, Nagda, Sevig, Thompson, & Dey, 1995). Although this research rarely, if ever, contains measures of students' interactions, interaction with diverse peers is a required component of such programs and is consequently assumed to be a factor in producing the outcomes.
One of the notable aspects of the work by Gurin et al. (2002; also see Gurin, 1999) is that both the quality and quantity of students' interactions with diverse peers are used. Their results suggest that the amount of interaction as well as the quality of that interaction influence outcomes such as active thinking and citizenship engagement. In an earlier study on the data used for our current analysis, the quality of students' interactions, particularly the positive quality, significantly predicted the importance students placed on social action engagement (see Hurtado, Nelson Laird, Landreman, Engberg, & Fernandez, 2002). In addition, that analysis identified relatively high positive correlations between the amount and quality of students' interactions with diverse peers. For these reasons, our current analysis focuses specifically on the positive quality of students' interactions with diverse peers.
Researchers have identified that students' predispositions tend to be accentuated in college (Feldman & Newcomb, 1969). Students intentionally [End Page 453] choose which college to attend, develop peer groups based on mutual interests, and enroll in courses that are interesting to them, resulting in the accentuation of particular attitudes, beliefs, and experiences. Astin (1993b), for example, found that gender differences on a range of attitudes tend to be accentuated—not diminished—during the college years.
The principle of accentuation does not apply, however, to every student in every setting (e.g., institutions, residences, courses, and peer groups). Because students do not and cannot always choose settings that "fit" their initial attitudes and values, it is likely that there are times when their predispositions will be challenged:
What the phenomenon of accentuation of initially prominent attributes reveals, then, must be stated conditionally: if students initially having certain characteristics choose a certain setting (a college, a major, a peer group) in which those characteristics are prized and nurtured, accentuation of such characteristics is likely to occur.(Feldman & Newcomb, 1969, p. 335)
Diversity courses are often considered a place where student attitudes and values will be challenged. Feldman and Newcomb's (1969) notion of accentuation reminds us that, at least for some students, a diversity course will serve as a mechanism for solidifying and extending characteristics they had upon entry into the course. Modeling and controlling for the accentuation of initial characteristics is, consequently, important in the study of diversity course effects.
Based on our review of the literature, the hypothesized model (see Figure 1) illustrates our understanding of the connections between enrollment in diversity courses, positive interactions with diverse peers, and the importance students place on social action engagement. At Time 1 (the beginning of an academic term), the number of previous diversity courses taken by students has a direct and indirect effect (through the amount of positive interaction students report having with diverse peers) on the importance students place on social action engagement. This relationship is repeated at Time 2 (the end of an academic term), with students' enrollment in a diversity course taking the place of the number of previous diversity courses. The existence of a direct effect is supported by the previous research connecting diversity courses and democratic outcomes (e.g., Gurin et al., 2002; Hurtado, Engberg, et al., 2002). The existence of the indirect effect relies on work that suggests that diversity programs directly influence the amount of interaction students report [End Page 454] having with diverse peers (e.g., Hurtado, Dey, & Treviño, 1994), and the research that suggests that the quality of interactions with diverse peers affects democratic outcomes (e.g., Gurin et al., 2002).
Two different accentuation effects are tested in the model. The first effect is an accentuation of the importance students place on social action engagement. We hypothesize that there is a significant indirect effect of the importance students place on social action engagement at Time 1 on the same measure at Time 2 through students' enrollment in a diversity course. The existence of such an effect would suggest that students predisposed to social action engagement would strengthen this attitude through their course enrollment. The second is the similar effect modeled for the amount of positive interaction students report with diverse peers. We hypothesize that students with higher scores at Time 1 will be more likely to enroll in a diversity course and that enrollment in a diversity course will in turn strengthen scores on each variable. Controlling for students' Time 1 scores (the inclusion of the direct path from students' Time 1 score to their Time 2 score on each variable) allows us to assess whether there is an effect of enrolling in a diversity course that is not accounted for by differences in scores at Time 1.
The last path in the model is the path from the number of previous diversity courses to the enrollment status in one of the study's diversity courses. This path is included to control for students' course-taking behaviors. By including this path, we are hypothesizing that students who [End Page 455] have taken diversity courses in the past are more likely to enroll in such courses in subsequent terms. This type of course-taking behavior is what can lead to strong accentuation effects over time.
The data in this study came from the Student Thinking and Interaction Survey (STIS) of college students, which was developed as part of a larger national research project titled Preparing College Students for a Diverse Democracy, conducted by Sylvia Hurtado. The survey was designed to assess students' cognitive and social development in the classroom over one term, with an emphasis on the mediating effect of students' interactions with diverse peers. The courses examined in this study were located at a flagship university in the Northeast, which was chosen for its interest in implementing additional classroom-based studies and assessing the impact of diversity courses.
Two types of courses are compared in this study: diversity and traditional courses. Data were gathered from students enrolled in two separate diversity courses and an introductory management course. The diversity courses, a social diversity course and a women's studies course, had a combined enrollment approximately equal to the enrollment in the management course. Both diversity courses met the campus-wide diversity requirement. In addition, the content in both courses was relatively similar (especially compared to the management course), and both courses provided in-class opportunities for interactions with diverse peers.
Students' responses to several questions about the pedagogy used in the courses (these questions were on the post-test) provide evidence of the similarities between the two diversity courses and the differences between the diversity courses and the management course. For example, 81% of the management students somewhat or strongly disagreed whereas 71% of the women's studies students and 98% of the students in the social diversity course somewhat or strongly agreed that group activities in class contributed to their learning. In addition, 72% of management students indicated somewhat or strongly agreeing that they had few opportunities to interact with classmates. Only 28% of the women's studies students and 8% of the social diversity students agreed with the same statement.
As these percentages indicate, the diversity courses frequently used active learning techniques (e.g., small group activities) that encouraged peer interaction across race. The management course did not provide [End Page 456] nearly as much opportunity for small group work or peer interaction. In addition, the diversity courses addressed several aspects of diversity (e.g., race, sex, and class) and several forms of oppression (e.g., racism, sexism, and classism) within the content of the course, while the management course rarely addressed issues of diversity. To further describe aspects of the different courses, a brief description of each course is given below.
The course on social diversity is a multiple section undergraduate course offered through the School of Education. One or more faculty members oversee the course while graduate students teach the actual sections of the course. Each section has approximately 25 students who are predominantly in their second or third year of college. The enrollment in each section is managed to some degree to encourage diversity among students' backgrounds and points of view. As a result, about 60% of the students in each section are usually female and about 20–25% are students of color.
The content of the social diversity course focuses on several forms of oppression prevalent in American society (e.g., racism and sexism). Students are expected to engage course material as well as to interact with one another throughout the semester. Readings and course activities are designed intentionally to foster such engagement. Instructors and students work to create a safe environment conducive to dialogue and learning. Students are expected to develop an understanding of how the different forms of oppression covered in the course operate in society and within themselves as well as to develop skills for exploring and counteracting multiple forms of oppression.
The women's studies course meets twice per week: once as an entire class (generally around 300 students) for a lecture or video presentation and once in smaller (around 20 students) discussion sections. A faculty member from women's studies oversees the course and has the primary responsibility for the lecture sessions while graduate students facilitate the discussion sections. As with most women's studies courses, there are generally many more women students (around 80%) in this course. Although enrollments are not intentionally managed in the same manner as in the education course, students of color make up about 20% of the students. For the most part, students are in their first and second year.
The women's studies course is intended to introduce students to concepts and perspectives basic to the field of women's studies. Through the examination of women's history and contemporary social issues, the course teaches students to read and think critically about many aspects of gender as well as about how issues of gender interact with issues of race, class, and sexual orientation. Learning through the discussion [End Page 457] sections is emphasized. Consequently, students in the women's studies course cover several types of oppression, and they are given time during the course to engage course material and issues with their peers.
The management course is a large lecture course designed to introduce students (generally about 350) to basic concepts and issues in management. The course meets twice per week, but a relatively high number of students are absent on any given day (a fact that affected our response rate for the management course). Students have reading assignments for each class, and exams are given in-class several times during the term. Like the education course, students are mostly in their second and third year. In contrast to the other two courses, the management course does not infuse issues of diversity throughout the course. There is, however, at least one day devoted to race and gender in management. In addition, there are few, if any, specific opportunities for students to interact with one another. The management course was selected specifically because its content and pedagogy were different from the other courses and reflected what can be identified as a more "traditional" approach. In addition, it is a relatively common course for students to take, and similar courses are found on many campuses around the country.
Instructors and students volunteered to participate during the spring 2001 semester and each faculty member gave homework points to encourage participation. In consultation with instructors and teaching assistants, it was determined to distribute the survey during class time, although students completed the survey on their own time. The STIS was administered to students during the 2nd week of class and again in the 12th week of the semester. At the beginning of the term, 311 out of 363 (85.7%) students in one of the diversity courses and 193 out of 345 (55.9%) in the management course filled out the STIS. The response rate for the diversity courses dropped to 73.6% at the end of the term while it remained the same for the management course. Not all of the same students filled out the STIS at the beginning and the end of the term. In total, 398 students (255 from the diversity courses and 143 from the management course) filled out the survey at both time points.
Of the 398 students who completed the instrument at the beginning and end of the term, 367 students (227 from the diversity courses and 140 from the management course) remained in the sample after listwise deletion based on all the variables used in the analyses. Of the student respondents from the diversity courses, 78% were first- and second-year students, whereas 22% were at least in their third year. The diversity [End Page 458] course respondents were predominantly female (80%) and White (75%). Of the 25% students of color, 8% were Asian, 8% were African American, 3% were Latina/o, and 7% were multiracial. The mean self-reported high school grade point average for those students in the diversity courses was 3.4 (based on a standard 4-point scale). Most respondents in the diversity courses (62%) indicated that their primary reason for enrolling was to meet a general education requirement (11% took the course to fulfill a requirement of a major or minor, and 27% took the course as an elective).
Student respondents in the management course were also predominantly White (76%). However, most of the respondents of color were Asian (15%), and smaller percentages were African American (2%), Latina/o (1%), or multiracial (6%). In contrast to the diversity courses, 52% of the management respondents were male and only 41% were in their first or second year. In addition, respondents from the management course overwhelmingly (87%) indicated that they enrolled in the course to meet the requirements of their major or minor. Lastly, respondents from the management course, on average, had approximately the same high school grade point average (3.3) as their counterparts in the diversity courses.
In order to evaluate the influence of a diversity course on the importance students place on social action engagement, two latent constructs (positive quality of interaction and social action engagement, each measured at the beginning and end of the term) and two observed variables (previous diversity courses and enrolled in a diversity course) were used in developing a structural model. A brief description of these variables follows. A full description of all observed variables used in the model is presented in the Appendix. 1
The social action engagement construct was composed of seven items that measure the importance students place on creating social awareness, volunteering for a cause, and working to eliminate poverty. Students were asked to rate each item on a 4-point scale from "not important" to "essential." Previous studies (see Hurtado, Engberg et al., 2002) suggest that these items represented a single construct that could be tested by confirmatory factor procedures. In order to create more stable parameter estimates and a more optimal variable to sample size ratio, items were categorized into 2- and 3-item parcels (see Bagozzi & Edwards, 1998; Bagozzi & Heatherton, 1994). Parceling is a proven technique to reduce bias in the estimates of structural parameters, and it results in better fitting solutions (Bandalos, 2002). 2 [End Page 459]
On the STIS, students were asked to identify the racial group other than their own with whom they interact most. For example, an Asian student in the sample might interact most with African American students, and an African American student might interact most with White students. Relative to interactions with that group, students were then asked to identify the frequency with which they had meaningful and honest discussions outside of class, shared personal feelings and problems, and worked effectively through conflict. Students rated each item on a 5-point scale from "not at all" to "a great deal." Student answers to these items were combined to make up the positive quality of interaction measure. Previous work connecting diversity experiences to educational outcomes has demonstrated the reliability of this scale and its usefulness in predicting democratic outcomes (Gurin, 1999; Gurin et al., 2002; Hurtado, Nelson Laird et al., 2002).
The diversity course enrollment variable was an endogenous, dichotomous variable that was coded 0 for students in the management course and 1 for those in a diversity course. The number of previous diversity courses taken by a student was an exogenous variable that indicates how many courses students took focusing on women's studies, ethnic/cultural studies, general diversity issues, intensive dialogue, and serving communities in need prior to enrolling in the diversity or management class. Students rated each class on a 4-point scale from "none" to "three or more," and these individual scored were added to get a measure of exposure to previous diversity courses.
In previous studies, exposure to diversity in the curriculum was measured by students' enrollment in either an ethnic studies course or a women's studies course (e.g., Astin, 1993a). Our previous diversity course measure, however, improves upon past measures by combining students' enrollment courses from ethnic and women's studies with courses that involve intensive dialogue across difference and serving a community in need. Although there may be legitimate concerns about defining all service-learning courses as diversity courses, at the particular institution where this study was based, the service-learning center is known to infuse racially/ethnically diverse content and methods into its courses. Saltmarsh and Heffernan (2000) also contend that service learning challenges students to engage actively in dialogues around issues of "equity, difference, inclusion, tolerance, justice, and power" (p. 5). Furthermore, many of the outcomes for service learning, such as the reduction of racial stereotypes and increased racial and cultural understanding (see Eyler & Giles, 1999; Astin & Sax, 1998), are consistent with outcomes based on more traditional diversity courses.
A previous study using the same data set found that, for the most part, background variables (e.g., gender, race, and grade point average) [End Page 460] indirectly influence variables measured at Time 2 through the corresponding measure at Time 1 (see Hurtado, Nelson Laird et al., 2002). The inclusion of the pretest measures in our model, consequently, serves as a control for the background characteristics; this explains why these variables were left out of the model. In addition, most background characteristics are single indicators, the inclusion of which could potentially create problems with model identification. Although beyond the scope of this study, future work on this data can address the effect of several background measures by examining the relative fit of our proposed model across different social identity groups.
The principal analytic technique used in this study consisted of confirmatory latent-variable structural modeling using the EQS 6 statistical software program (Bentler, 1995; Bentler & Wu, 2002). Structural equation modeling allows for the simultaneous estimation of hypothesized relationships using estimated covariance/correlation matrices and generates goodness of fit measures to evaluate the overall fit of the proposed model. Structural equation modeling provides several advantages over traditional path analysis, including the assessment of the overall fit of a hypothesized model and the ability to take into account measurement error to obtain more precise coefficient estimates.
Our analyses were based on a correlation matrix using listwise deletion for handling missing data. Although structural equation models are commonly constructed using covariance matrices, the inclusion of a categorical variable in our model (i.e., enrollment in a diversity course) necessitates the use of a correlation matrix. In order to better estimate the relationship between a categorical variable and a continuous variable, a polyserial correlation is used. There is not a comparable measure for covariances between categorical and continuous variables. Consequently, a matrix containing correlations and polyserial correlations is required to construct the model (see Bentler & Wu, 2002). Using this matrix in conjunction with maximum likelihood (ML) estimation under robust procedures results in good parameter estimates, adjusted and new test statistics, and corrected standard errors (Bentler & Wu, 2002). 3 Although our latent constructs (i.e., positive quality of interaction and social action engagement) contain variables that are not strictly continuous, the characteristics of these variables (e.g. number of categories, skewness, and kurtosis) fall within ranges where any distortion of parameter estimates will be small given our sample size and use of ML (see Bollen, 1989; Muthén & Kaplan, 1985). [End Page 461]
We relied on guidelines suggested by Raykov, Tomer, and Nessleroade (1991) and reported the goodness of fit measures known as the Normed Fit Index (NFI), Non-normed Fit Index (NNFI, which is also known as the Tucker-Lewis Index or TLI), and the Comparative Fit Index (CFI). Further, we followed Boomsma's (2000) recommendation to include the misfit index known as the Root Mean Square Error of Approximation (RMSEA). Current standards for determining acceptable fit suggest that fit indices should exceed .90 and RMSEA scores should be equal to or below .10. Models considered to fit well have fit indices in excess of .95 and RMSEA scores equal to or below .05. We also included the Satorra-Bentler scaled chi-square and the Yuan-Bentler residual-based F-statistic, which are recommended for structural models that use categorical variables (Bentler & Wu, 2002). To reduce the sensitivity of the chi-square statistic, we divided its value by the degrees of freedom. Although there are no clear-cut guidelines about what value is minimally acceptable, Kline (1998) suggests that the ratio be less than 3. Klem (2002) further supports the use of this measure, especially for smaller sample sizes like the one in this study.
We began our analyses by using confirmatory factor analysis to test our measurement model and to determine how well observed variables and their corresponding latent constructs fit the data. At this stage, we tested the factor structure of two factors (positive quality of interactions and social action engagement) at two time periods (4 factors in all).
Next, we tested the full structural model and examined the corresponding goodness of fit and RMSEA indices. In accordance with the principles of modeling longitudinal data (see Kline, 1998), we included measurement error correlations between each pair of Time 1–Time 2 observed variables. Additionally, we imposed equality constraints on the factor loadings of each repeated measure. We examined both the Wald and the Lagrange Multiplier tests to determine whether any modifications were needed, and we found no theoretical basis to make any changes. Finally, we examined the significance of each of the structural paths as well as data on the direct, indirect, and total effects for all of the variables in the model.
Several factors place limits on the conclusions and generalizations that can be drawn from this study. The first limit is our reliance on self-reported data to model change in students' dispositions to engage in social action. While there are many disadvantages to using self-reported data for this purpose (e.g., social desirability of answers can influence [End Page 462] results), the higher education research community currently lacks widely used objective measures of civic engagement. Rather than relying solely on administrative assumptions about how students develop these skills, administrators have a viable alternative in self-report data for informing decision-making in higher education (Pike, 1995). Subsequent work with these data is also planned to explore the link between our own survey measures and more widely adopted measures on standardized instruments (e.g., Defining Issues Test of moral and ethical development).
Second, our choice to locate this study within particular diversity courses and a comparison course limits the ability to generalize from the results. The results are best understood as applying to the women's studies course and the social diversity course in comparison to the management course. However, the similarity of these results to previous studies (e.g., Gurin et al., 2002) suggests that a body of evidence is being accumulated that demonstrates the effectiveness of diversity experiences in promoting democratic outcomes across contexts and in comparison to several alternative types of experiences.
Third, apart from syllabi and what the students reported about their own classroom experiences, limited data were gathered on the design and daily operations of the courses involved in this study. We lack, therefore, a qualitative understanding of what happened in each course to produce the results reported below. In order to build on the results of this study, we recommend that future work of a similar nature use a mixed method approach (employing both quantitative and qualitative methods) to further our understanding.
Fourth, the sample of students who took the STIS at the beginning and end of the term was smaller than ideal for the analytic methods that were employed. This limited the number of variables that could be included in the model. However, small sample sizes are an inherent problem in conducting quantitative research at the classroom level, and, compared to most classroom studies, our sample size did not exclude us from using a multivariate technique such as structural equation modeling.
The results for the measurement model suggest a good fit between the data and our two latent constructs (quality of interaction and social action engagement). The chi-square statistic is 97.45 (df = 46, n = 377, p < .001), and the fit indices are NFI = .96, NNFI = .97, CFI = .98, and RMSEA = .06. Table 1 shows the standardized parameter estimates (factor loadings) for each observed variable that was regressed on its [End Page 463] respective latent construct. Additionally, Table 1 contains the r-square for each observed variable to latent construct relationship, which indicates the amount of variance explained in each factor by its corresponding observed variable. Freely estimated parameters were all highly significant (p< .001), confirming the hypothesized factor structures and providing an empirical justification for proceeding to our full structural model.
Table 2 shows the parameter estimates and significance levels for each of the direct and indirect paths in the structural model. Figure 2 provides a more complete schematic that includes the paths for errors and disturbances as well as the correlations among error terms over time. Results indicate that the proposed structural model fits the student data quite well. The chi-square and Satorra-Bentler scaled chi-square are 142.92 and 125.54 (df = 66, n=367, p< .001), respectively, which indicates that χ2 / df is between 1.90 and 2.17. The fit indices are NFI = .94, NNFI = .96, CFI = .97, and RMSEA = .05. Thus, our model presents an appropriate and meaningful depiction of how students' exposure to diversity courses influences the amount of positive interaction students report having with diverse peers and the importance they place on social action engagement.
At Time 1, the number of previous diversity courses students took produced a significant direct effect (b = .21, p < .001) on students' social action engagement. In addition, the number of previous diversity [End Page 464]
courses students took was a significant predictor (b = .27, p < .001) of students' positive quality of interactions, which was in turn a significant predictor (b = .19, p < .01) of social action engagement. The resulting indirect effect of the number of previous diversity courses on social action engagement (Time 1) was also significant (b = .05, p < .01). This suggests that prior exposure to diversity courses predisposes students to have higher amounts of positive interactions with diverse peers and to place higher value on the importance of social action engagement. [End Page 465]
As expected, the strongest paths in the model were between Time 1 and Time 2 measures of the positive quality of interaction and social action engagement, the repeated measures, indicating students' predispositions account for a large share of the variance in the end-of-term measures. Even with these controls, however, students' enrollment in a diversity course had significant positive effects on both positive quality of interaction (Time 2) and social action engagement (Time 2). This suggests that even over the course of a single term, the diversity courses [End Page 466] (compared to the management course) increase the amount of positive interactions students report with diverse peers and increase the importance students place on taking social action.
After controlling for scores at Time 1 and enrollment in a diversity course, the path from positive quality of interactions to social action engagement was not significant at Time 2 (b = .06, p > .05). We know that there is a significant zero-order correlation between positive quality of interactions and importance of social action engagement at Time 2 (r = .25, p < .001). The nonsignificance of the path in the model suggests that the variance explained by positive quality of interactions is also explained by enrollment in a diversity course, the importance of social action engagement at Time 1, or both. It seems likely, since the diversity courses in this study explicitly promoted positive interactions across race, that the explained variance is shared by the enrollment in a diversity course variable.
The number of previous diversity courses did not significantly predict current enrollment in a diversity course (b = .07, p > .05). Since the courses included in the study were offered at an introductory level, our results suggest that prior enrollment in diversity courses does not predict future enrollment in introductory diversity courses. The hypothesized effect of prior diversity courses may exist, however, for enrollment in upper-level diversity courses.
Enrollment in a diversity course directly affects (b = .17, p < .01) the amount of positive interaction students report having with diverse peers at Time 2. However, there is no significant direct effect of positive interactions at Time 1 on enrollment in a diversity course (b = .02, p > .05). Enrolling in a diversity course, as a result, does not appear to accentuate students' entering level of interactions with diverse peers, as we initially hypothesized.
With regard to the other accentuation effect, students' predisposition toward social action engagement did produce a significant direct effect (b = .19, p < .001) on their choice to enroll in a diversity course. Enrollment in a diversity course, in turn, had a significant direct effect (b = .15, p < .05) on social action engagement at Time 2. Consequently, diversity courses appear to accentuate the importance students place on social action engagement.
Overall, our model accounts for 57% of the variance of the importance of social action engagement and 42% of the variance of positive quality of interactions at the end of the term. Although pretest scores account for much of the variance in both variables, enrollment in a diversity course rather than the management course contributes significantly in both cases. Given the differences between the two types of courses, this suggests that the inclusion of diverse content and diverse methods of [End Page 467] instruction encourages the development of students who are more socially aware and socially active. However, the limited nature of this study (a comparison between two diversity courses and a management course) requires that more research is needed to confirm this result.
This study makes several important contributions to our understanding of how diversity courses can influence students' interactions with diverse peers and the importance they place on social action engagement. The results demonstrate that previous enrollment in diversity courses and enrollment in one of the diversity courses in this study (compared to enrollment in a management course) are positive, significant determinants of the quality of students' interactions with diverse peers. As such, diversity courses in general and the particular diversity courses in this study are preparing students for the inherent challenges that await them as they enter an increasingly diverse workforce. Bikson and Law (1994), for instance, report that the business community is currently looking for future workers with the skills to work effectively in groups with colleagues of diverse backgrounds. Many diversity courses, including the two in this study, are working to achieve this goal by teaching students skills that translate into interchanges with diverse peers that are more open and meaningful. Additionally, these courses teach students important skills in negotiating and communicating across difference—skills necessary in a workforce that values collaboration and dialogue among diverse team members.
This study also shows that previous enrollment in diversity courses and enrollment in one of the diversity courses in the study positively influenced students' commitment to social action engagement, an outcome that indicates students' desire to take actions in their communities and relationships in order to end social injustices. In addition to the direct effects, the results suggest that the number of previous diversity courses increases the importance students place on social action engagement indirectly through the amount of positive quality of interactions students have with diverse peers. The indirect effect of enrollment in one of the diversity courses was not significant at Time 2, however, because positive quality of interactions did not explain a significant amount of variance in social action engagement independent of the other variables in the model, even though the correlation between the measures indicates a significant relationship.
These results seem especially important given the growing public disaffection toward civic engagement (Putnam, 2000) and students' excessive individualism and lack of accountability toward one another (Halstead, 1999). Barber (1997) insists that creating responsible citizens is [End Page 468] a task "colleges and universities can be expected to undertake, for it reflects nothing more than a recognition of and recommitment to the traditional ideal of education as preparation of young people for civic life in a free society" (p. 228). According to Newman (1985), the crisis in education is less about declining test scores and more about the failure "to provide the education for citizenship that is still the most important responsibility of the nation's schools and colleges" (p. 7). Furthermore, Checkoway (2001) suggests that the decline in civic engagement is embedded in the multiple systems and levels of many institutions of higher learning, including classroom pedagogy that no longer incorporates the development of civic competencies. Thus, diversity courses like those in this study seem an appropriate and available educational solution for institutions interested in improving their students' commitment to the larger public good.
Two separate accentuation effects (Feldman & Newcomb, 1969) were investigated in this study. The results suggest that enrollment in a diversity course (vs. enrollment in a management course) does accentuate the importance students place on social action engagement. In other words, students who were more committed to social action engagement tended to take a diversity course, which in turn strengthened this attitude. The results, however, do not support the assertion that enrollment in a diversity course accentuated the amount of positive interactions students report having with diverse peers. Unlike social action engagement, higher amounts of positive quality interactions did not make it more likely that a student would enroll in one of the diversity courses. This suggests that for the students in this study, the amount of positive interactions they were having with diverse peers did not substantially affect the likelihood that they would enroll in one of the diversity courses or the management course. However, given that this study was on three particular courses, further study is needed to draw conclusions about the existence of this accentuation effect. In addition, the concept of positive interactions with diverse peers is a broad one that could be measured in several ways and by drawing from a larger set of items. The use of another measure of positive quality of interactions may confirm the existence of an accentuation effect.
This study provides evidence of a significant effect for enrollment in a diversity course on positive interactions with diverse peers and social action engagement (even after controlling for students' prior exposure and experiences). Although the inclusion of both knowledge and support for diversity is evident in ethnic and women's studies programs, cocurricular academic support programs, and multicultural programming (Muñoz, 1989, Peterson et al., 1978; Treviño, 1992), this study and others like it could provide the empirical support for the continued inclusion of diverse perspectives and methods of instruction in classrooms across campus to encourage the development of the behaviors [End Page 469] and attitudes needed for civic engagement. The accentuation effect demonstrated in this study points to the continued learning students can glean from courses that include diverse content and structured interactions with diverse peers.
To build on the study reported here, our future research will examine more closely the development of democratic outcomes, a phrase used here to encompass the dispositions, knowledge, and skills necessary for active participation in our pluralistic society, in courses that focus on diverse content versus those courses that include both structured interactions among diverse peers and diverse content. In addition, studies will examine the efficacy of the model reported here in predicting other democratic outcomes (e.g., perspective taking) within classroom-based studies as well as within larger longitudinal samples from multiple institutions. Beyond the research to be done on the Preparing College Students for a Diverse Democracy project, additional research is needed on different diversity courses, along with a wide variety of comparison courses. The results of such work will likely speak to the necessity of a continued emphasis on curricular transformation within higher education and will help determine which particular elements of a course (e.g., purpose, content, teaching methods) need to change in order to promote increased civic engagement among students.
Implications for Research and Practice
The existence of an accentuation effect for the importance students place on social action engagement confirms our belief that controlling for the possibility of these effects is critical to work done at the classroom level, especially given that much of students' course-taking likely serves an accentuation function. Structural equation modeling was well suited for our analysis, especially in light of our interest in examining and controlling for accentuation affects. Although there are some limitations to using structural equation modeling (e.g., the inclusion of numerous variables is more problematic than in regression analyses), we feel it has proven to be a powerful technique for examining classroom data with multiple time points. Unlike using regression analyses, we were able to account for indirect effects. In addition, we were able to examine measurement error and the overall fit of our model, both of which are strong advantages over regression and path analysis. Researchers with suitable data at the classroom level should consider using structural equation modeling for their future analyses.
While the college curriculum has become more representative of America's cultural legacies over the last three decades, "the college peer group [End Page 470] constitutes the place where students gain experience, experiment, and learn how to negotiate differences in backgrounds and perspectives that are an inevitable part of contemporary society" (Hurtado, 1999, p. 26). As educators, we need to continue to explore ways to take advantage of the potential for learning from peers. The results of this study add to a growing body of research that suggests that the creation of classroom structures, content, and pedagogy that encourage multicultural understanding and that maximize the benefits of positive interactions with diverse peers are important components of realizing the power of peer-to-peer learning for civic commitment (e.g., see Adams & Zhou-McGovern, 1994; Chang, 2002; Gurin et al., 2002; Hurtado, 2001a). However, such pedagogical change requires action by faculty and instructors. It is likely that many instructors will need assistance in making these changes, especially in preparing to facilitate the dialogues that transpire within an increasingly diverse campus community. If these changes promote the cultivation of citizens and leaders who are more engaged in social action, then higher education will be fulfilling its role in society as instigator of progress.
1. For interested readers, a table containing means and standard deviations for all the observed variables in the model and a table containing the covariance matrix for the variables in the model are available upon request from the authors.
2. For readers who wish to learn more: Item parceling is a proven technique discussed in detail by Bagozzi & Edwards (1998), Bagozzi & Heatherton (1994), and Bandalos (2002) and treated generally by Kline (1998). A major concern in the use of parceling is that item parcels may mask the multidimensionality of the full item set (see Hall, Snell, & Singer Foust, 1999). In this study, we have confidence, based on prior research (e.g., Hurtado, Engberg et al., 2002; Hurtado, Nelson Laird et al., 2002), that social action engagement is a unidimensional construct. That fact combined with the benefits of parceling compelled us to use the technique.
3. For comparison purposes, the model was also run using covariances without changing the enrollment in a diversity course variable to be categorical. The fit indices and most of the parameter estimates are very close, except for several path coefficients involving the enrollment in a diversity course variable. Most notably, the coefficients for the paths leading from the diversity course variable to Factor 3 and Factor 4 are smaller in the model based on covariances, suggesting that the attenuation that can be caused by the inclusion of a categorical variable has been corrected in the model based on the polyserial correlations.