Student Experiences with Diversity at Liberal Arts Colleges: Another Claim for Distinctiveness

This study explores the relationship between organizational and individual characteristics and diversity-related experiences at liberal arts colleges. Compared with their counterparts at other types of institutions, students at liberal arts colleges report more experiences with diversity. In addition, this study also finds that diversity experiences are positively related with a variety of student outcomes.

diversity experiences show greater relative gains in critical and active thinking (Gurin, 1999;Pascarella, Edison, Nora, Hagedorn, & Terenzini, 1996). Experience with diversity also appears to be positively associated with retention rates and degree aspirations (Chang, 1999), more frequent participation in community service (Bowen & Bok, 1998;Gurin, 1999), and higher levels of civic engagement, cultural awareness, and commitment to improving racial understanding (Milem, 1994). Finally, diversity experiences seem to favorably influence overall satisfaction with the college experience and perceptions of the campus climate (Astin, 1993;Chang, 1999Chang, ,2001Milem & Hakuta, 2000).
This brief summary suggests not only that diversity experiences are related to desired substantive outcomes of college, but also that they shape the way students think about themselves in relation to others, about the nature of the activities in which they engage, and about the value they place on attitudes toward others and their skills and competencies in working with different types of people during and after college (Gurin, 1999). That is, as a result of experiencing diversity in college, students learn how to work effectively with others and how to participate actively and contribute to a democratic society. Moreover, through engaging with people from different backgrounds and with different life experiences, students are adding to the foundation of skills and dispositions that is essential for living a productive, satisfying life after college in an increasingly multicultural world. Thus, the very act of experiencing diversity during college helps students develop the habits of the mind and heart that enlarge their capacity for doing so after college (Kuh, 2003;Shulman, 2002).

Forms of Diversity Present on College Campuses
Three forms of diversity that potentially influence the way students think and behave exist to varying degrees on college and university campuses (Chang, 2001;Chang, 2002;Gurin, 1999;Milem & Hakuta, 2000). The first, structural diversity refers to the numerical representation of students from different racial and ethnic groups in the student body (Hurtado, Milem, Clayton-Pederson, & Allen, 1998;Hutado et aI., 1999). The greater the number of students from different backgrounds, the more likely it is that a student will have an opportunity to interact with someone from a different race or ethnic background. In addition, a critical mass of students of color is important in creating an environment that uses diversity to enhance learning processes.
A second type of diversity involves the number and nature of diversityrelated initiatives that colleges and universities make available. These include among others required multicultural or diversity courses (often as part of the general education component), elective ethnic studies courses, cultural awareness workshops offered during orientation and throughout the year, and cultural centers. While demographic shifts or changes in the structural diversity of campuses frequently stimulate the creation of these efforts (Chang, 2001), some colleges and universities make these opportunities primarily because their campus lacks racial and ethnic diversity (Kuh, Kinzie, Schuh, Whitt & Associates, 2005).
The third form of diversity, diversity interactions, is represented by students' exchanges with racially and ethnically diverse people as well as their exposure to diverse ideas, information, and experiences. Social psychologists claim that the more interaction one has with others who hold different views, or the more one learns about various aspects of human diversity, the more likely it is that one will be challenged to think and respond in novel ways. For example, people who interact with more complex social structures exhibit a heightened sense of individuality while simultaneously showing a more complex attentiveness to the social world (Coser, 1975). Gurin (1999) contends that racially and ethnically diverse learning environments provide the types of complex social structures that stimulate the development of active thinking processes.
These three forms of diversity are not mutually exclusive. In fact, students often learn the most about diversity in its many forms not from classes or workshops but through interacting with others whose views and cultures differ from their own. Moreover, while diversity-related initiatives benefit students who are exposed to them-even on campuses that are racially and ethnically homogeneous-their impact on students can be more powerful on campuses that have greater structural diversity (Chang, 2002). In sum, while each form of diversity can confer significant positive effects on educational outcomes, the impact of each is multiplied by being exposed to the other types of diversity (Change, 1999(Change, , 2001(Change, , 2002Gurin, 1999;Hurtado et al., 1998Hurtado et al., , 1999. Conversely, the impact of each form of diversity is diminished in settings where the other forms are less prevalent (Milem & Hakuta, 2000).

What Other Institutional Conditions Foster Experiences With Diversity?
The evidence of the impact of diversity on student learning and development is very promising. At the same time, we know relatively little about the factors and institutional conditions that promote and enhance students' experiences with diversity. Virtually all types of colleges and universities assert the value and promise of exposing students during college to all forms of human diversity (Alger, 1997). But do the benefits of diversity extend equally to students in all types of institutions? Hu and Kuh (2003) found that students in private institutions more frequently interacted with students from different backgrounds and that students at large doctoral-extensive universities and liberal arts colleges had more experiences with diversity than their counterparts at other types of institutions had. It is not surprising that students at large universities would have more exposure to diversity, given that these institutions typically enroll more students from different racial, ethnic, and cultural groups. Somewhat unexpected is that students at smaller liberal arts colleges would report equally frequent experiences with diversity. Historically, small liberal arts colleges have claimed to have distinctive missions, especially when compared with large public universities (Clark, 1970;Kuh, Schuh, Whitt, & Associates, 1991;Townsend, Newell, & Wiese, 1992). But they also tend to be located in rural and less racially diverse locations. Even so, it appears that a distinctive dimension of contemporary liberal arts colleges is their ability to expose students to diversity in educationally purposeful ways. How they do this is not clear.

Purpose
This study examines the nature of student experiences with diversity at liberal arts colleges. Specifically, three questions guide the investigation.
1. How do students at liberal arts colleges compare with their counterparts at other types of institutions in terms of their diversity-related experiences? 2. What organizational factors and conditions are related to diversity experiences at liberal arts colleges? 3. What are the relationships between diversity experiences at liberal arts colleges and other educationally purposeful activities and outcomes?

Data Source
Data for this study come from the National Survey of Student Engagement. The NSSE is an annual survey of first-year and senior students that measures the degree to which students participate in educational practices that prior research shows are linked to valued outcomes of college (Chickering & Gamson, 1987;Kuh, 2001;Pascarella & Terenzini, 2005).
The NSSE survey instrument, The College Student Report, asks students about their experiences in four areas: (a) the amount of time and effort devoted to various in-class and out-of-class activities, (b) participation in enriching educational activities (e.g., study abroad, intern-ships), (c) gains in personal and educational development, and (d) perceptions of the college environment, including overall satisfaction with college and quality of academic advising. The survey relies on student self-reports. A fair amount of research (Baird, 1976;Berdie, 1971;Pace, 1985;Pike, 1995;Pohlmann, 1974) has shown that self-reports are likely to be valid if (a) the information requested is known to the respondents, (b) the questions are phrased clearly and unambiguously, (c) the questions refer to recent activities, (d) the respondents think the questions merit a serious and thoughtful response, and (e) answering the questions does not threaten, embarrass, or violate the privacy of the respondent or encourage the respondent to respond in socially desirable ways . The NSSE survey was designed to satisfy all of these conditions.
To answer the guiding research questions we draw on two overlapping samples of students. The first sample was composed of 98,744 undergraduates (49,706 first-year students, 49,038 seniors) from 349 4-year colleges and universities who responded to the NSSE survey in spring 2002. The second sample is a subset of the larger sample and is composed of 17,640 (9,598 first-year students, 8,042 seniors) undergraduate students enrolled at 68 Baccalaureate Colleges-Liberal Arts, I as defined in the Carnegie classification scheme (McCormick 2001).

Data Analysis
The dependent variables were measures of student engagement, perceptions of the campus environment including satisfaction, and selected self-reported outcomes of college (Appendix A). Student engagement was measured using four scales: (a) level of academic challenge, (b) classroom activities that represent higher order thinking (a subset of items that contribute to the academic challenge measure), (c) active and collaborative learning, and (d) diversity-related activities. The supportive campus environment measures included four measures: a supportive campus environment scale, two subscales (interpersonal support and support for learning) of supportive campus environment, and an overall satisfaction with college scale. Students' gains in learning and intellectual development were represented by two scales: gains in general education and gains in personal and social development. In addition, three individual measures were used to represent gains in social awareness: gains in understanding self, gains in desire to contribute to community, and gains in understanding others. The items that make up the gains scales are student self-reports about the extent to which their college experience has contributed to their growth in these areas.
The data were analyzed in three stages using hierarchical linear modeling (HLM). Because of the nested nature of the data and the intent to estimate institutional effects (Raudenbush & Bryk, 2002), we used HLM. When conducting research on organizational effects, researchers are faced with a decision about unit of analysis. Should they build regression models by aggregating to the institution level, or should they attach institution-level characteristics to individuals? If researchers build models at the institution level, they are prone to the "ecological fallacy," whereby individual differences are masked (King, 1999;Kreft & de Leeuw, 1998). For example, an analysis based on colleges might reveal that students at smaller colleges are more engaged in effective educational practices than are students at large colleges, while an analysis of small colleges might reveal that many students at large colleges are as engaged or more engaged than are small-college students.
Research where institution-level characteristics are attached to an individual is flawed as well (Ethington, 1997). First, it violates the general assumption of ordinary least squares (OLS) regression: Observations are independent of one another. Second, it assumes that individuals within a group are affected identically by group-level characteristics. Finally, the inclusion of group-level variables into an OLS regression equation leads to misestimation of standard errors and to the wrong number of degrees of freedom, increasing the likelihood of committing a Type I error (indicating that something is statistically significant when it is not).
Using HLM overcomes the problems associated with unit of analysis by simultaneously modeling both individual and institutional effects. HLM partitions the variance between the institution and the student, resulting in more accurate parameter estimates. This makes it possible to determine what is an individual-level effect or a group-level effect. Because these effects can be partitioned, each can be modeled with their respective characteristics.
First, we used HLM to explore the impact of differences in student engagement in diversity-related activities by institutional type. In HLM, we are able to allow the intercept to vary, thereby partitioning the variance between the institution and the student. We then are able to model the average institution differences using institutional characteristics. At the institution level, we created dummy coded variables for the five collapsed Carnegie institutional categories for 4-year colleges and universities: Doctoral/Research-Extensive, Doctoral/Research-Intensive, Master's I & II, Baccalaureate Colleges-Liberal Arts, and Baccalaureate Colleges-General.? To determine whether diversity experiences of students at Liberal Arts Colleges differed from those of their counterparts at other types of institutions, Liberal Arts Colleges was designated as the omitted    group. To control for possible confounding variables we included several student-level variables in all of our models. Separate models were run for 1st-year students and seniors. Table 1 presents the descriptive statistics for the independent variables included in the models.
In the second stage, we used HLM to examine the institutional factors that are related to diversity and the effects of diversity experiences on student engagement at liberal arts colleges. Again, we allowed the intercept to vary allowing us to determine the variance that can be attributed to individual student differences. In addition to student-level controls, we included several institution-level controls such as sector (public, private), urbanicity (urban, suburban, small town/rural), size (total undergraduate headcount), and a measure of selectivity (derived from the 2002 Barron's College Guide). We modeled separately three institutional measures of different aspects of diversity. First, we tested the influence of structural diversity using a variable that represents the probability that a student will interact with a student from another race (Meyer & McIntosh, 1992). 3 We call this the "diversity density index." We selected this measure because it most accurately represents the amount of diversity on a given campus. As Chang (1999) notes, many studies make the assumption that more students of color equals a more diverse student body and use the proportion of non-White students as their measure of structural diversity. However, this measure is flawed because it does not take into account the heterogeneity of some college campuses. For example, a Historically Black College or University (HBCU) would be considered highly diverse given that nearly all of the students on campus are non-White. If our fundamental rationale for the benefits accrued by diversity is that students learn from those who are different from them, students at HBCUs may be less likely to reap such benefits. Our "diversity density" measure overcomes this heterogeneity problem by including all five racial/ethnic groups in an equation. If the percentages of the five racial/ethnic groups on a given campus is nearly equal (e.g., 20%, 20%, 20%, 25%, 15%), then it is more likely that a student will interact with someone from a different race than on a campus where the percentages of the five groups is varies greatly (e.g., 75%, 5%, 10%, 10%,0%).
Second, we tested the effects of "climate for diversity," or students' perceptions (institutional average) of the emphasis their institution places on encouraging contact among students from different economic, social, and racial or ethnic backgrounds. Third, we tested the influence of "diversity in coursework," the extent to which students reported their classes included readings or discussions related to diversity. Finally, we modeled a construct which we call "diversity press," by creating a scale made up of the three other diversity measures (senior ex = .71; first-year student a =.66). Diversity press represents the institution's commitment to and emphasis on diversity as manifested by the proportional presence of students from different backgrounds attending the institution (structural diversity), the extent to which students perceive that diversity is valued and important, and the degree to which diversity is featured in the curriculum.
In the third and final stage of the analysis, we built a series of hierarchical linear models to explore the relationships between students' engagement with diversity-related activities at liberal arts colleges and measures of student engagement in educationally purposeful activities and their self-reported gains in personal and educational growth. We also examine the impact of diversity-related activities on perceptions of student environment and satisfaction.
We calculated effect sizes to assess the magnitude of differences between students' experiences with diversity at liberal arts colleges and the types of institutions to help interpret the results of the regression equations (Rosenthal & Rosnow, 1991). The effect size is the proportion of a standard deviation change in the dependent variable as a result of a oneunit change in an independent variable. We standardized all of the continuous independent and dependent measures in the models, so the unstandardized coefficients represent effect sizes. The larger the effect size the more likely the differences between groups represent performance that warrants serious discussion and, perhaps, intervention. Taking the advice of Rosenthal and Rosnow (1991), we considered an effect size of .10 or less to reflect a trivial difference, between .10 and .30 small, between .30 and .50 moderate, and greater than .50 large. Table 2 presents the coefficients for 1st-year students and seniors from the multilevel models comparing liberal arts colleges to other types of institutions on average engagement in diversity-related experiences and self-reported gains in understanding people from diverse backgrounds. Both first-year and senior students at liberal arts colleges, on average, are more likely than their peers at other types of colleges and universities to engage in diversity-related activities (effect sizes ranging from .19 to .30). While the effects are somewhat smaller (.10 to .23), students at liberal arts colleges reported statistically significantly higher gains in understanding diverse people than did students at other types of colleges.

Institutional Context
The high levels of student engagement in diversity-related activities and self-reported gains in understanding people from different back- Level I controls-age. race. gender, transfer. grades, Greek. major, full-time, first generation college grounds at liberal arts colleges suggest that many of these colleges offer a distinctive educational experience for students compared with those offered by other types of schools. The next set of models is run only on students from liberal arts colleges and attempts to explain why these institutions perform so well compared with others by exploring the influence of their institutional features on diversity-related experiences. Table 3 presents the coefficients for the four institutional diversity measures. The diversity density index, or the probability that a student will interact with a student of another race, significantly predicts student engagement and gains in only a few areas. As expected, first-year and senior students at institutions with high diversity density scores are more likely to engage in diversity-related activities and report gains in understanding people from other backgrounds and cultures. Diversity density also is positively related to gains in personal and social development for seniors, but it is negatively related to satisfaction and interpersonal support.
The institutional climate for diversity, or encouragement to interact with people from different backgrounds, is positively related to almost all of the engagement and gains measures. As expected, the strongest effects for both first-year and senior students can be seen in the average engagement in diversity-related activities and gains in understanding diversity. However, the degree to which an institution encourages interactions with diverse peers is not significantly related to level of academic challenge for either seniors or first-year students. Diversity in coursework, or the frequency with which students encounter diverse perspectives in class, is also positively related to many of the dependent measures. However, for both first-year and senior students, overall satisfaction with college and the supportive campus environment measure were unrelated to emphasizing diverse perspectives in the classroom.
Given these findings, it is no surprise that the diversity press construct is positively related to many of our dependent measures. As with the results from the other models in Table 3, diversity press is most strongly related to engagement in diversity-related activities and gains in understanding diversity for both first-year students and seniors. For first-year students, diversity press was positively related to all the dependent measures except academic challenge, satisfaction, and supportive campus environment. For seniors, the only measures that were not related to diversity press were satisfaction, supportive campus environment, and gains in understanding others. Table 4 displays the effect sizes for engaging in diversity-related activities after controlling for institutional and individual characteristics  for students at liberal arts colleges. Student experiences with diversity at liberal arts colleges are positively related to all the dependent measures. Students who participate in diversity-related activities report higher levels of academic challenge, participate more frequently in active and collaborative learning, report greater gains in personal and educational growth, and report greater satisfaction with their college experience. They also perceive that their campus environment more strongly supports their academic and social needs. The effect sizes range from .27 to .44, with the largest effects related to academic challenge, active and collaborative learning, and personal and social development gains.

Limitations
Because this study focuses primarily on institutional effects, we used HLM. However, the between-school variance is quite small for many of the dependent measures. The intraclass correlations (ICC), or the proportion of variance explained by institution membership, ranged from .04 to .11. Some might argue that it is unnecessary to use multilevel models with such small ICCs. At the same time, there are two compelling reasons why the HLM approach is preferred over OLS regression for this study. First, all of the model reliabi1ities are quite high (none are below .6), which suggests that using multilevel modeling to estimate the intercepts is appropriate (Kreft & Leeow, 1998). Second, if we attempt to model group-level effects with OLS by attaching institutional characteristics to individuals, we will not accurately capture the effect of these characteristics and will misestimate their standard errors (Ethington, 1997).
Another limitation is related to the validity of self-reported gains. As Pascarella (2001) and others point out, gain scores may be confounded by students' entering characteristics. However, Pike (1999) provides some evidence to suggest that gain scores are not significantly related to entering ability. Although the concerns about self-reported data are legitimate, the gains measures are only one of several sets of dependent variables used in this study. In addition, because of ceiling effects, we might expect the self-reported gains of students at liberal arts colleges, many of which employ more selective admissions standards, to be smaller overall than those of students at master's-granting and baccalaureate general colleges. That is, students at the latter types of schools might start college at lower levels of intellectual and personal development and might have more room to move upward on the self-report gains measures. Inasmuch as the pattern of results overwhelmingly favors liberal arts colleges, the concerns about the validity of self-reported gains should be interpreted in the context of the complete set of findings.
Our results should also be considered in light of possible inadequacies in the 2000 Carnegie Classification scheme. Some might argue that the system is so diverse even within categories in the characteristics of institutions that the categories themselves have little meaning. While we recognize such limitations, the Carnegie Classification is the most recent and best known way to categorize colleges and universities.
Additionally, some of the effect sizes are relatively small. In particular, some of the effects of the institutional diversity measures for liberal arts colleges are trivial. Even so, the pattern of the effect sizes and the magnitude of other effect sizes favoring liberal arts colleges cannot be overlooked.
Finally, given that the data used in this study are cross-sectional, we are unable to control for self-selection. Perhaps students who matriculate to liberal arts colleges are more predisposed to seek out people from different backgrounds or believe that these colleges provide them with opportunities for these types of interactions. While the latter is not likely, given the limited structural diversity of many liberal arts colleges, the former is possible and warrants a cautionary note.

Discussion
The findings from this study suggest that liberal arts colleges create distinctive learning environments for students in terms of diversity experiences. The advantage is consistent across the outcomes measures used in this study and often supported by substantial effect sizes even after controlling for student and other institutional features. Students at liberal arts colleges are significantly more likely than are their counterparts at other types of institutions to engage in diversity-related activities and to report greater gains in understanding people from diverse backgrounds.
In terms of the overall effects of diversity experiences across different types of colleges and universities, the results generally corroborate the research showing positive relationships between diversity and desirable outcomes of college. That is, students who engaged in diversity-related activities more frequently reported higher levels of academic challenge, greater opportunities for active and collaborative learning, and a more supportive campus environment. They were also more satisfied with their college experience and reported greater gains in a variety of areas since starting college.
Liberal arts colleges generally perform well on measures of student engagement compared to other types of colleges and universities (Astin, 1993(Astin, , 1999Hu & Kuh, 2003;Kuh, 2003;Kuh & Hu, 2001;National Survery of Student Engagement, 2000Pascarella, Wolniak, Cruce, & Blaich, 2004;Pascarella & Terenzini, 2005). But the pattern of results favoring liberal arts colleges in terms of diversity experiences is surprising, if for no other reason that many of these institutions are not naturally imbued with some forms of diversity. For example, many liberal arts colleges are located for historical reasons in rural settings (Rudolph, 1990), which are not racially or socioeconomically diverse or which are not viewed as desirable collegiate environments by many students from historically underrepresented groups (e.g, students of color, first-generation college students). As a result, liberal arts colleges should not be expected to score well on the diversity density index. Figure 1 displays the relatively low structural diversity at liberal arts colleges. Although we see differences between and within institution types, only baccalaureate-general institutions had a lower median diversity density index than liberal arts colleges.
Consistent with some previous research, first-year students' diversity density was negatively related to satisfaction and perceived interpersonal support (Hu & Kuh, 2003;Rothman, Lipset, & Nevitte, 2003). However, the diversity press model does not show these negative effects, indicating that the negative impact of diversity density on satisfaction and interpersonal supported is ameliorated when institutions offer courses that emphasize diverse perspectives and create a climate that promotes interac- tions between people from different backgrounds. This finding is consistent with Hurtado et al. (1999) and Chang (1999), who argue that structural diversity in and of itself does not necessarily result in an environment supportive of diversity. In order for institutions to optimize the benefits of structural diversity, they need to display a commitment to diversity and provide opportunities for diverse interactions (Chang, 1999, p. 392). Our findings support these claims and suggest that one way to do this is to present diverse perspectives in the classroom. Additionally, institutions need to find other ways to communicate the value of diversity and to support the academic and social needs of students from different backgrounds. In the context of liberal arts colleges, most of which tend to be relatively small, apparently the magnitude of the number of students from different backgrounds does not matter as much to deriving the benefits of diversity experiences as does the quality of interactions across differences that the campus environment encourages and nurtures.

Implications
Taken together, the findings from this study indicate that an institution does not have to be highly structurally diverse to foster meaningful diversity experiences. Few of the liberal arts colleges in this study were as or more structurally diverse than, for example, larger research-intensive universities. Yet, liberal arts colleges appeared to have created environments that took advantage of the limited diversity on campus.
Institutions that use the absence of structural diversity as an explanation for why they are unable to enhance student learning through diversityrelated activities could look to liberal arts colleges for ideas as to how they are able to promote diversity experiences.
Liberal arts colleges are not the only type of institutions that have effectively infused diversity experiences into the undergraduate program. In fact, much can be learned from large universities such as the University of Michigan, University of Maryland, Indiana University, and Arizona State University about their innovative approaches that facilitate cross-group interactions. These institutions have implemented intergroup dialogues that bring together diverse groups of students with the purpose of discussing issues related to their diversity. So far, the research on the impact of these experiences is promising, suggesting that these dialogues are an effective way for students to learn to become citizens in an increasingly diverse society (Schoen & Hurtado,200 I).
The field would benefit from additional efforts to codify programs, policies, and practices both at liberal arts colleges, which as a group seem to be relatively effective at promoting beneficial experiences with diversity, and at other types of institutions that perform better than predicted in this area (Garcia et al., 2002). In addition, institutional audit tools similar to those described by Green (1989) and Kuh, Kinzie, Schuh, and Witt (in press) that assess the factors and conditions that encourage students to experience diversity could be profitably used by institutions and to external groups such as accreditation organizations.

Conclusion
This study adds to our understanding of institutional characteristics that are related to students' experiences with diversity. As colleges and universities prepare students to live and work in an increasingly diverse democracy, many liberal arts colleges appear to have created learning environments distinguished by an accentuated diversity press. Opportunities to interact across racial, religious, and socioeconomic lines seem to be the critical factor in terms of creating an environment that supports experiences with diversity and for cultivating the skills and dispositions that are considered essential to be effective in an increasingly multicultural society. Although structural diversity is not a sufficient condition for realizing the benefits that accrue from experiencing diversity, recruiting students from a broad range of backgrounds, including race and ethnicity, remains important. Indeed, without a minimal level of structural diversity, it would be very difficult to have the ingredients in place to promote the cross-cultural interactions that are essential for deep learning about diversity.