FACTORS INFLUENCING POSITIVE INTERACTIONS ACROSS RACE FOR AFRICAN AMERICAN, ASIAN AMERICAN, LATINO, AND WHITE COLLEGE STUDENTS

Research in Higher Education, Vol. 48, No. 1, February 2007 (Ó 2006) DOI: 10.1007/s11162-006-9026-3 FACTORS INFLUENCING POSITIVE INTERACTIONS ACROSS RACE FOR AFRICAN AMERICAN, ASIAN AMERICAN, LATINO, AND WHITE COLLEGE STUDENTS Victor B. Saenz,*, ** Hoi Ning Ngai,* and Sylvia Hurtado*... This study explores the various factors that promote positive interactions across race for African American, Asian American, Latino, and White college students. A longitudinal survey was administered to all incoming students at nine public institutions (with a follow-up survey given at the end of their second year), examining activities related to cross-racial interaction and outcomes. This knowledge will be useful for college administrators, institutional researchers, and faculty as they work to meet the challenge of preparing students for a pluralistic society. The results of this study begin to delineate the conditions under which positive intergroup relations can be fostered in college and the key factors that inhibit or restrain the benefits such interactions bring to the development of students for a diverse workplace and pluralistic democracy.... KEY WORDS: diversity; affirmative action; cross-racial interactions; intergroup relations; intergroup contact; admissions; race-conscious policies. INTRODUCTION Rapid demographic changes within the United States and increased demands for access to postsecondary education have compelled institutions to continue diversifying their student bodies and focus on improving intergroup relations. While some associate greater diversity on college campuses with increased student interactions across race and positive educational outcomes (Astin, 1993; Chang, 1996; Antonio, *Higher Education Research Institute, Los Angeles Graduate School of Education and Information Studies, University of California, Box 951521, Los Angeles, CA, USA. **Address correspondence to: Victor B. Saenz, Higher Education Research Institute, Los Angeles Graduate School of Education and Information Studies, University of California, Box 951521, Los Angeles, CA 90095-1521, USA. E-mail: vsaenz@ucla.edu 1 0361-0365/07/0200-0001/0 Ó 2006 Springer Science+Business Media, Inc.

2 SAENZ, NGAI, AND HURTADO 1998), others allege that it produces fewer interactions, more self-segregation, and balkanized college environments (D Souza, 1991; Rothman, Lipset, and Nevitte, 2003). In fact, some critics have gone so far as to assert that greater student diversity stems out of misguided affirmative action policies, provoking the development of ethnocentrism (D Souza, 1991) and the aggravation of racial tensions among students (Thernstrom and Thernstrom, 1997). Underlying this ongoing debate is whether it is simply enough to have a diverse population of students or whether it is more imperative for students of different backgrounds to be connecting with each other in meaningful ways. While previous research (Chang, 1996; Chang, Astin, and Kim, 2004) focuses on the ability of diverse campus environments to provide more opportunities for cross-racial contact, current research articulates a more nuanced argument. Recent studies elaborate on the importance of quality interactions, indicating that greater racial and ethnic diversity on campuses can enhance learning environments if opportunities for positive and meaningful interactions are available for students (Bowen and Bok, 1998; Hurtado, Milem, Clayton-Pederson, and Allen, 1999; Antonio, 2001; Gurin, Dey, Hurtado, and Gurin, 2002). This shift in research focus suggests a need to reconsider not only the extent of cross-racial contact, but, more importantly, the nature of the interactions that take place. This study directly addresses this need, by exploring factors that predict positive cross-racial interactions among college students. Furthermore, this research aims to provide insights into the student- and institutional-level conditions necessary for achieving these positive intergroup interactions. BACKGROUND Across multiple arenas, the debate over affirmative action has shifted from a discussion of moral and compensatory issues toward an understanding of the value of diversity within organizations and societyat-large (Milem, 2001). In the higher education setting, affirmative action in the form of race-conscious admissions policies has come under increasing scrutiny as a result of statewide initiatives (e.g., Prop. 209 in California), Appellate Court decisions (e.g., Hopwood v. Texas), policy responses (e.g., Talented Twenty Percent in Florida), and two recent U.S. Supreme Court decisions (i.e., Grutter v. Bollinger, 2003, and Gratz v. Bollinger, 2003). While we face a more varied landscape of institutional practices in terms of how greater racial and ethnic diversity can be achieved on college campuses, the reasons for why higher education must continue to

POSITIVE INTERACTIONS ACROSS RACE 3 diversify have become more focused. In particular, the recent Supreme Court decisions were notably influenced by a prodigious amount of empirical research that elaborated on the educational benefits of diverse learning environments. However, in affirming that the goal of increasing diversity serves a compelling educational interest, the Grutter and Gratz decisions also instituted a narrow framework to restrict how institutions can use race-conscious admissions polices to achieve this goal. While the moral and compensatory arguments for increasing racial and ethnic diversity have not been abandoned, scholars must also focus on building a stronger empirical case for the educational value of diversity in higher education. This argument cannot be effectively made, however, without understanding better the conditions that lead to improved cross-racial interactions among college students, interactions that are necessary for achieving important educational outcomes that can be derived from more diverse institutional settings. LITERATURE REVIEW Within the past decade, a number of studies have argued for the importance of racial and ethnic diversity on college campuses, especially in developing life-long knowledge and skills. Astin (1993) found that students who engaged in diversity activities (e.g., discussing racial issues, socializing with someone of another race, attending racial/ cultural awareness workshops, taking ethnic or women s studies courses) reported greater gains on cultural awareness and commitment to racial understanding. Chang, Hakuta, Jones, and Witt (2003) articulated numerous benefits that diverse campuses facilitated for students, including higher-order thinking skills, less racial stereotyping, and greater likelihood of living, working, and socializing comfortably in integrated settings. Other studies have probed the issue further by considering how diversity brings about educational outcomes and directing attention to crossracial contact as a possible mechanism. In linking campus diversity with student interactions, Chang (1996) found that more racially diverse campuses had a positive, direct effect on the frequency in which students engaged in cross-racial interaction and discussed issues of race. When these interactions take place, empirical studies have suggested that they play an important role in achieving educational benefits. Students who reported frequent interactions with diverse peers showed a greater openness to diverse perspectives and a willingness to challenge their own beliefs after the first year of college (Pascarella, Edison, Nora, Hagedorn, and Terenzini, 1996). More recently, findings from a national report on ten public institutions revealed that positive interactions

4 SAENZ, NGAI, AND HURTADO across race were significantly correlated with various cognitive, social, and democratic outcomes (Preparing College Students for a Diverse Democracy, Hurtado, 2003). Research has examined the effects of cross-racial interactions in both formal and informal college contexts. Formal interaction with others, through diversified curricula and diversity courses, had a consistently positive effect on complex thinking skills, retention, cultural awareness, interest in social issues, and support for institutional diversity initiatives (Chang, 2001; Hurtado, 2003). At the informal level, interactional diversity (i.e., the extent of interaction with diverse others) was observed to positively impact interpersonal and leadership skills (Antonio, 2000; Hu and Kuh, 2003). In a recent study by Saenz (2005) that employed a similar outcome measure and dataset as this study, students who reported a greater extent of diverse informal interactions prior to arriving in college were likely to report more frequent positive interactions during college and were more likely to have complex views of racial discrimination, as compared to their counterparts with fewer pre-college interactions. These results indicate a strong empirical link between students formal and informal cross-racial interactions both before and during college and important college outcomes related to diversity efforts. In addition to evaluating contexts of interaction, prior research has also addressed patterns of contact across racial/ethnic groups. Mack et al. (1995) found that different groups vary in their comfort level when interacting with others based on the contexts in which those interactions occur. For example, Asian students indicated the greatest discomfort when interacting with other groups in academic or social environments, while African American students indicated discomfort particularly when interacting with White students in more intimate contexts. In contrast, both Latino and White students expressed little discomfort when coming into contact with diverse others. A recent study by Chang et al. (2004) found that increasing the diversity of the student body had differential effects on patterns of cross-racial contact among White students as compared to students of color. While the pattern for White students suggested that the degree of cross-racial interaction was a function of the availability of students from other racial/ethnic groups, an explanation of the pattern for students of color was not as apparent, warranting additional investigation in future studies. Although research on diversity has been important in establishing the positive effects of cross-racial interaction, many of these studies were limited by their inability to pinpoint the conditions that influence the character of students experiences with diversity, relying on measures of frequency rather than quality. As findings from Chang et al. (2004) and

POSITIVE INTERACTIONS ACROSS RACE 5 Mack et al. (1995) suggested, students of varying racial backgrounds often experience diversity in complex ways, indicating a need for a more detailed analysis of those factors that might promote and facilitate positive cross-racial interactions. Saenz (2005) offered an important venture into this research void with his focus on the saliency of pre-college measures in affecting students positive cross-racial interactions during college. In the interest of extending this area of research as well as addressing the shortcomings of prior work that did not fully account for students racial backgrounds, we posed the following research question: What individual characteristics (including social psychological predispositions) and environmental conditions (including campus climate and institutional practices) promote positive interactions across race for African American, Latino, Asian American and White students in the second year of college? By adding to the existing body of research on the potential educational benefits of diversity, this study sought to advance our understanding of the student- and institutional-level conditions that predict positive cross-racial interactions in the early college years. Through the use of longitudinal data collected at college entry and at the end of the second year of college, we focused particularly on the influence of individual attitudes, curricular and co-curricular contexts, and various peer environments. Ultimately, we anticipate that our findings can inform the ongoing debate over the dynamics of racial and ethnic relations in higher education by offering empirical evidence on the ways that institutions can better facilitate positive cross-racial interactions, thereby enhancing the overall learning environment and campus climate for all students. CONCEPTUAL FRAMEWORK Given the extent to which prior research already provides strong support for the positive relationship between cross-racial interactions and educational outcomes, the motivating force behind this study is to determine the conditions for promoting positive intergroup relations or interactions on campus. Likewise, the focus is not on simply increasing the frequency of intergroup contact, but rather on improving the quality of interactions. If we are searching for ways in which students can engage in positive interactions with diverse others, what contexts and experiences are most likely to provide meaningful opportunities for dialogue and exchange? What factors inhibit substantive intergroup contact? Student self-reports of their positive interactions across race served as the principal outcome measure in this study, as we sought to uncover those factors that most effectively facilitated these exchanges on college campuses.

6 SAENZ, NGAI, AND HURTADO In order to examine the impact of these various factors, we utilized Dovidio et al.õs (2004) model of intergroup bias to provide key concepts with which to develop the conceptual framework for this study. Within their model, the authors address the extent to which enlightenment (i.e., acquisition of knowledge) and contact (i.e., interaction with others) work together through mediating processes (i.e., intervening affective aspects of the social environment) to affect intergroup bias. Instead of assessing the level of intergroup attitudes, as the model was designed to do, we chose instead to apply the concepts of enlightenment and mediating processes to the development of a conceptual framework to predict the quality and substance of intergroup contact on campus. For our purposes, we considered enlightenment and contact at two distinct points within our model: enlightenment serving as a category for a number of independent variables reflecting formal and normative opportunities for students to learn about diverse people and quality of contact serving as the dependent variable. Within this study, we used the term enlightenment to refer to those contexts that allowed for campusfacilitated exposure to diversity content and ideas (e.g., curriculum integration, extracurricular programs, intergroup dialogue in courses) and the term contact to refer to meaningful interactions with diverse others. In addition to considering the effect of enlightenment on contact, we also attempted to examine the role of mediating processes surrounding enlightenment and to determine how these processes influence students choice of experiences as well as their decision to engage with diverse others. We accounted for these mediating influences in two areas of the framework. First, we acknowledged the role that cognitive and affective processes can play with regard to a student s choice of enlightenment experiences. We posited that these mediating influences include an individual s level of attributional complexity (cognitive), level of intergroup anxiety (affective), and predisposition to learn about diversity (attitudinal). Ultimately, these qualities reflect a student s mental and emotional readiness for exposure to diversity in college. Next, we recognize that additional mediating processes can impact whether or not particular college experiences translate into meaningful contact with diverse peers. These mediating influences include: (1) a student s sense of connection with his/her respective institution, a connection that often stems out of his/her perception of the campus climate for racial/ethnic diversity, and (2) a student s propensity to socialize, a quality that may compel some students to take a more active approach towards engaging with diverse peers. The notion is that students feeling marginalized or who lack general social skills are least likely to have positive interactions across

POSITIVE INTERACTIONS ACROSS RACE 7 Support for Diversity Edu. Student Characteristics Pre-College Environment Attributional Complexity Intergroup Anxiety Campus- Facilitated Diversity Initiatives Peer Groups Campus Climate Propensity to Socialize POSITIVE INTERACTIONS ACROSS RACE Quality of Contact Pre-College Characteristics Pre-College Mediating Processes Opportunities for Enlightenment (in College) Comfort in Social Environment (in College) racial/ethnic groups, or perhaps other social groups at their particular college. The diagram in Fig. 1 offers a visual representation of the conceptual framework, taking into account pre-college experiences, college environments, and mediating processes on positive interactions across race. In the following review of our framework components, we articulate our rationale for why we chose particular variables to include in the study. Pre-college Considerations FIG. 1. Conceptual framework. Pre-college environments and interactions with diverse others tend to be highly correlated with interactions across race/ethnicity during the undergraduate years (Saenz, 2005; Hurtado, Engberg, Ponjuan, and Landreman, 2002). Depending on previous exposure to diverse people in their pre-college environments, students may be more or less inclined to interact with students of other racial/ethnic backgrounds upon arriving on college campuses (Saenz, 2005). Moreover, the extent to which students meaningfully engage others from different backgrounds in high school can determine their levels of attributional complexity (i.e., how they think about others behavior) and intergroup anxiety (i.e., how comfortable they feel with diverse people) upon college entry. Attributional complexity is a term derived from Fletcher, Danilovics, Fernandez, Peterson, and Reeder s (1986) work on the ability of individuals to attribute behaviors beyond immediate group contexts (such as those defined by race) and move past generalizations and stereotypes. Intergroup anxiety is a concept taken from the work of Stephan and Stephan (1985), examining varying levels of comfort and tension across racial/ethnic groups. These competencies, along with a predisposition toward learning about diversity, serve as mediating influences on whether students choose to engage in particular enlightenment experiences.

8 SAENZ, NGAI, AND HURTADO College Experiences Allport (1954) notes that one of the necessary conditions for positive intergroup contact and reduction of prejudice is the existence of authority or institutional support. Taking this into account, we examined the impact of campus-facilitated diversity initiatives, including curricular and co-curricular activities. Research into diversity coursework and activities suggest that such experiences allow students of different backgrounds to expand their knowledge of themselves and others through opportunities to discuss difficult issues and express different opinions (Chang, 1996; Hurtado, 1998; Zuniga, Nagada, and Sevig, 2002). In addition to diversity coursework, service learning, participation in intergroup dialogue, and other co-curricular activities, we also considered academic support services and faculty encouragement of student discussions as ways in which institutions can guide students toward more meaningful interactions across race. Although various structured experiences can promote positive interactions within academic spaces, additional research suggests that meaningful contact also occurs in informal settings, specifically those in which students work, study, and socialize together. These informal interactions are often determined by students choices in peer groups, including those defined by fraternities/sororities, ethnic organizations, athletics, and student government. Allport (1954) suggests that intimate relationships are the ones that are the most effective in allowing individuals to cross racial/ethnic boundaries and to learn from each other in meaningful ways. In these situations, where anxieties are attenuated, more opportunities may exist for conversation and dialogue to be positive and transformational. Another significant determinant of positive contact with diverse peers is overall comfort within the campus environment. Pettigrew (1998) notes that the context for intergroup contact must provide participants with the opportunity to become friends, suggesting that a supportive environment is necessary for quality interactions to take place. Levin, Van Laar, and Sidanius (2003) found that negative perceptions of campus climate led members of different racial/ethnic groups to have more ingroup friends rather than expand to other groups. As such, student perceptions of campus climate can have a strong impact on whether interactions take place at all. Finally, we accounted for those students who are simply more comfortable in social settings and may be more prone to interacting with others regardless of racial/ethnic backgrounds. We acknowledged these qualities of comfort as key mediating influences on whether students are able to bridge racial/ethnic differences. These

POSITIVE INTERACTIONS ACROSS RACE 9 are important variables to consider, as those who feel more connected and comfortable within the campus social environment whether it is due to their participation in campus diversity initiatives or their connection to peer groups may experience higher levels of positive contact. METHOD Data Source This study utilized longitudinal data from the Preparing Students for a Diverse Democracy Project, a multi-campus national research project that surveyed students at college entry and again at the end of their second year. The Diverse Democracy Project was aimed at providing empirical evidence to inform the debate over the educational benefits of diversity and offer guidance for institutional practice that would link teaching and learning with diversity (Hurtado, 2003). Institutions were invited to participate in the Diverse Democracy Project because they exhibited a strong commitment to diversity initiatives as exemplified through curricular and co-curricular programming and because they had demonstrated recent success in diversifying their campuses (2003). Students were surveyed at college entry in the Fall of 2000 and again at the end of their second year in the Spring of 2002. Survey participation was restricted to students who matriculated in the Fall of 2000 at one of the 10 participating institutions. After one of the campuses was dropped from the longitudinal study due to an extremely low follow-up survey response rate, the remaining nine campuses had a return rate of about 36% (n = 13,520) for the first-year survey and a return rate of 35% (n = 4,757) for the second-year survey (Hurtado, 2003). Ultimately, the sample for this study included 4,757 students from nine public universities, with all students completing both the first-year and second-year surveys. To correct for low response rates and to generalize our results to the original sample population, statistical weights were created by institution to account for response bias in each survey (2003). Weights were then readjusted to ensure that the new weighted sample did not produce incorrect standard errors leading to possible Type I errors. Further, a review of the dataset revealed a small range of missing data (0 10%) for each of the measures in our analyses. In order to retain as much of the sample as possible, missing values for all independent measures were replaced using the EM algorithm function in the SPSS statistical software, a useful method for replacing missing data when a small proportion of a dataset is missing (2003). Once weighting was

10 SAENZ, NGAI, AND HURTADO completed and missing data were replaced, the final sample consisted of 4,380 students distributed across the four racial/ethnic groups of interest: 686 Asian students, 224 African American students, 388 Latino students, and 3,082 White students. Measures and Analyses The dependent measure for our study positive interactions across race was constituted using a scale of 7 items (alpha = 0.88). This measure consisted of items that assessed the extent to which one engaged students of other racial/ethnic backgrounds in: dining/sharing meals, having discussions about racial/ethnic relations outside of class, sharing personal feelings/problems, studying/preparing for class, socializing/partying, having intellectual discussions outside of class, and attending events sponsored by other racial/ethnic groups. The theme of community and sharing across these items suggests a positive and meaningful character to the interactions being assessed. The individual factor loadings ranged from 0.52 to 0.85 for the entire sample, and the factor held together well across all four racial groups under investigation, with the Cronbach s alpha value ranging from 0.85 for African Americans to 0.89 for Latinos. A total of 22 pre-college and college measures were selected for the analyses, including: 14 single-item variables, 7 factor scales, and 1 institutional-level measure. The independent variables and the factor scales were chosen based on a review of those items that fit into our conceptual framework as well as those that have been found to be significant in prior research conducted as part of the Diverse Democracy Project. Variable definitions are shown in Table 1, along with coding schemes and scales. Items constituting factors are shown in Table 2. Factor analyses, using principal axis factoring and varimax rotation, were conducted to confirm the saliency of the outcome measure and of the other factors employed in our analyses. Factor scales were created using the regression method, which centers and standardizes the scale for each factor, preserving the variance of each individual item while enhancing the overall utility of the newly created scale. Additionally, an analysis of variance and post hoc t-tests were conducted to assess whether the four racial/ethnic groups were statistically different from each other on the outcome measure. Once all measures were thoroughly inspected and confirmed through missing data and factor analyses, separate multivariate regression analyses were run for each of the four racial/ethnic groups, comparing the unstandardized beta coefficients for each of the groups to determine the strongest predictors for positive interactions across race. Unstandardized

POSITIVE INTERACTIONS ACROSS RACE 11 TABLE 1. Variable Definitions and Coding Scheme Dependent Variable Positive Interactions Across Race Seven-item Standardized Factor Scale* Independent Variables Background Characteristics SAT (math/verbal) Continuous: math: 200 800; verbal: 200 800 Gender (Female) Dichotomous: 1 = male; 2 = female. Socioeconomic status (family income, mother s/father s highest level of education) Pre-college Environment Pre-college racial environment (neighborhood, high school, friends) Extent of interactions with individuals from different racial/ethnic backgrounds in high school Frequency of studying with individuals from different racial/ethnic backgrounds in high school Three-item factor scale Three-item factor scale Count: Interaction w/asians, Blacks, Latinos, Native Americans, Whites; four-point scale: 1 = no interaction, to 4 = substantial interaction Five-point scale: 1 = never, to 5 = daily Pre-college Mediating Processes Pre-college support for diversity education Five-item factor scale Pre-college level of attributional complexity Five-item factor scale Pre-college level of intergroup anxiety (frequency of Count: Discomfort w/asians, Blacks, Latinos, discomfort with different groups) Native Americans, Whites; four-point scale: 1 = never, to 4 = often

12 SAENZ, NGAI, AND HURTADO TABLE 1. (Continued) Institutional Characteristics Structural diversity (percentage of underrepresented minorities) Continuous: 0 100 Peer Groups Lived on campus Dichotomous: 1 = no, 2 = yes Participated in leadership training Dichotomous: 1 = no, 2 = yes Participated in Greek organization Dichotomous: 1 = no, 2 = yes Joined organization(s) that promote cultural diversity Dichotomous: 1 = no, 2 = yes Campus-Facilitated Diversity Initiatives Participated in diversity co-curricular activities Two-item factor scale Participated in diversity course(s) Three-item factor scale Had opportunities for intensive dialogue between Four-point scale: 1 = none, to 4 = three or more students of different backgrounds in classroom(s) Participated in service learning course(s) Four-point scale: 1 = none, to 4 = three or more Took advantage of academic support services (tutoring, academic advising) Two-item factor scale Faculty demonstrated interest in development Four-point scale: 1 = strongly disagree, to 4 = strongly agree College Mediating Processes Perception of racial tension Four-point scale: 1 = strongly disagree, to 4 = strongly agree Hours per week socializing Six-point scale: 1 = 0 hours, to 6 = over 20 hours *Note: All factor scales were computed using standardized scales (regression method) unless otherwise noted. Blacks refers to African Americans.

POSITIVE INTERACTIONS ACROSS RACE 13 TABLE 2. Items Constituting Factor Scales Factors and Items Scales and Cronbach s Alphas Loadings Positive interactions across race Alpha =.88 Dined or shared meal 1 = never, 5 = very often 0.74 Had meaningful and honest discussions about 1 = never, 5 = very often 0.72 race/ethnic relations outside of class Shared personal feelings and problems 1 = never, 5 = very often 0.79 Studied or prepared for class 1 = never, 5 = very often 0.67 Socialized or partied 1 = never, 5 = very often 0.73 Had intellectual discussions outside of class 1 = never, 5 = very often 0.85 Attended events sponsored by 1 = never, 5 = very often 0.52 other racial/ ethnic groups Socioeconomic status Alpha =.69 Family income 1 = < $10,000, to 8 = > $150,000 0.51 Mother s highest level of education 1 = some high school, to 6 = doctorate or 0.66 professional degree (e.g., JD, MD, PhD) Father s highest level of education 1 = some high school, to 6 = doctorate or 0.86 professional degree (e.g., JD, MD, PhD). Pre-college racial environment Alpha =.84 Racial/ethnic composition of neighborhood 1 = all or nearly all people of color, to 0.85 5 = all or nearly all white Racial/ethnic composition of high school 1 = all or nearly all people of color, to 0.88 5 = all or nearly all white Racial/ethnic composition of friends in high school 1 = all or nearly all people of color, to 0.89 5 = all or nearly all white

14 SAENZ, NGAI, AND HURTADO TABLE 2. (Continued) Factors and Items Scales and Cronbach s Alphas Loadings Pre-college support for diversity education Alpha =.80 Incorporating writings and research about different 1 = strongly oppose, to 4 = strongly support 0.69 ethnic groups and women into courses Requiring students to complete a community-based 1 = strongly oppose, to 4 = strongly support 0.68 experience with diverse populations Offering courses to help students develop an 1 = strongly oppose, to 4 = strongly support 0.68 appreciation for their own and other cultures Requiring students to take at least one cultural or 1 = strongly oppose, to 4 = strongly support 0.72 ethnic diversity course in order to graduate Offering opportunities for intensive discussion between 1 = strongly oppose, to 4 = strongly support 0.63 students with different backgrounds and beliefs Pre-college level of attributional complexity Alpha =.78 I am interested in understanding how my own thinking 1 = not at all like me, to 5 = very much like me 0.59 works when I make judgments about people I really enjoy analyzing the reason or causes 1 = not at all like me, to 5 = very much like me 0.66 for people s behavior

POSITIVE INTERACTIONS ACROSS RACE 15 I think a lot about the influence that society has on other people I believe it is important to analyze and understand our own thinking processes I think a lot about the influence that society has on my behavior 1 = not at all like me, to 5 = very much like me 0.73 1 = not at all like me, to 5 = very much like me 0.64 1 = not at all like me, to 5 = very much like me 0.62 Participated in diversity co-curricular activities Alpha =.79 Campus-organized discussions on racial/ethnic issues 1 = never, to 5 = very often. 0.81 Diversity awareness workshops 1 = never, to 5 = very often 0.81 Enrolled in diversity courses Alpha =.85 Material/readings on gender issues 1 = none, to 4 = three or more 0.74 Material/readings on issues on oppression 1 = none, to 4 = three or more 0.83 Material/readings on race and ethnicity issues 1 = none, to 4 = three or more 0.84 Took advantage of academic support services Alpha =.69 Tutoring sessions for specific courses 1 = never, to 5 = very often 0.73 Academic support programs 1 = never, to 5 = very often 0.73

16 SAENZ, NGAI, AND HURTADO regression coefficients (b) were used for comparative purposes to evaluate betas across each of the four groups. These coefficients can be more easily interpreted to determine differences in the effect of a variable on one group versus another (i.e., an interaction effect). For example, comparing unstandardized regression coefficients (b) allows for conjectures to be made about how one specific pre-college or college experience has greater predictive strength for White students positive interactions than for the other three racial groups. Where large differences in betas were observed across racial groups, the coefficients for each of these groups were compared by way of a two-tailed t-test to determine whether the effect of a measure on one population was significantly different from its effect on another population. The critical p-value to determine t-test significance was set at p <.005. The t-tests (Sax, Bryant, and Harper, 2005) were conducted as follows: (b = beta coefficient; S b = standard error of b; 1 = first group; 2 = second group): t ¼ b 1 b 2 q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : S 2 b1 þ S2 b2 In order to examine the predictive power of our independent measures, we used blocked hierarchical regression, entering our variables in seven blocks based on temporal order of impact and as informed by our conceptual framework. The variables were placed into the following blocks: background characteristics, pre-college environment, pre-college mediating factors, institutional characteristics, peer groups, campusfacilitated diversity experiences, and college mediating factors. Conceptual/Analytic Model The first block of variables consisted of background characteristics: SAT scores, gender, and socioeconomic status. These items recognize prior research linking high-ability and high SES to greater likelihood of interactions with diverse others (Hurtado, Carter, and Sharp, 1995). The second block of variables in the regression model consisted of three measures of pre-college environment, one describing students racial environment (in their surroundings as well as in their friendship groups) and the other two measures describing the frequency and quality of interactions with diverse peers prior to college entry. Such measures serve as proxies that represent the extent of students pre-college exposure to diverse others, which often sets the tone for whether or not students feel comfortable engaging with diverse others in college. The third block included three measures of pre-college mediating processes:

POSITIVE INTERACTIONS ACROSS RACE 17 support for diversity education, level of attributional complexity, and level of intergroup anxiety. These measures provide a sense of how ready and willing students are to take in diverse perspectives and people from diverse backgrounds, potentially directing students toward particular experiences and outcomes. After accounting for student attitudes and experiences prior to college, the fourth block of variables consisted of a single institutional measure of structural diversity, as defined by the percentage of underrepresented minorities. In order to construct the item, percentages of Asian, African American, Latino, and Native American students were obtained for each of the institutions and summed to create the total percentage. The fifth block considered the role of peer groups in facilitating student experiences with diverse others. The block consisted of peer groups as defined by: living on campus, participating in leadership training, participating in Greek organizations, and joining organizations that promote cultural diversity. The sixth block of variables focused specifically on campus-facilitated diversity initiatives. The block included measures ranging from participation in diversity curricular and co-curricular activities, to faculty support of student development and interaction, to use of academic support services. The final block consisted of two additional factors that were considered college mediating processes. These processes reflect attitudes and abilities that may have overriding effects on the likelihood of positive interactions across race. The first measure, perception of racial tension, has been previously discussed in the context of students of color and their likelihood to self-segregate in racially tense environments (Levin et al., 2003). The second measure, hours per week socializing, acknowledges that those who are more naturally outgoing will be more likely to interact with diverse others and more capable of crossing racial/ethnic divides. Each of the seven blocks was force entered into the regression model for all four racial/ethnic groups. The effects of the 22 independent measures on the dependent measure, as well as the relationships between the independent variables, were assessed by examining the changes in their unstandardized regression coefficients (b) and the proportion of variance explained (adjusted R 2 ) over the course of the model. Limitations An important limitation of this study lies in the fact that the population of White students within our sample is substantially larger than the populations of Asian, African American, and Latino students. In conducting separate analyses for each of these racial/ethnic groups, it is

18 SAENZ, NGAI, AND HURTADO likely that many more independent variables were deemed statistically significant for the White students given the large sample size. To address this, we set the significance threshold at a higher level (i.e., p <.001) for this particular regression model. While the low response rate across students was another limitation, weighting processes were utilized to address this issue. The small number of institutions (n = 9) in this dataset necessitated more conservative assessments of the institution-level variables, and ultimately only one such measure (i.e., structural diversity or proportion of students of color) was employed in our regression models. Furthermore, the fact that all of the institutions in this dataset were public research universities suggests that the results of our analyses may not be necessarily generalizable to other 4-year public or private institutional settings. Nonetheless, the public universities represented in this study represent arenas where the debate about diversity and the question about intergroup relations is likely to arise. The value of these data and analyses can also speak volumes for the potential of empirical research to inform the conditions for intergroup relations that students will carry forward into post-college careers. RESULTS In comparing the mean-centered values (i.e., the overall sample mean was centered at zero, thereby creating a z-score for the measure) for the dependent measure across racial/ethnic groups, African American students reported the greatest levels of positive interactions across race (l=.33) at the end of the second college year, with White students reporting the lowest levels (l=).11), a figure well below the overall sample mean. African Americans, along with Latinos (l=.17) and Asians (l=.28), reported mean values that were well above the overall sample mean. Also, the standard deviation statistics indicate the least variability for African American students, thereby lending greater support to the idea that these students may be having more positive interactions with those of different racial/ethnic backgrounds on public university campuses. These initial trends across racial groups suggest important descriptive differences in students levels of positive interactions, a suggestion that is further examined through post hoc tests (see Table 3 for results). The Dunnett T3 post hoc test, which acknowledges heterogeneous variances across groups, indicates that significant differences in the dependent measure exist between White students and each of the other groups. That is, there is a statistically significant difference (p <.001) in mean

POSITIVE INTERACTIONS ACROSS RACE 19 TABLE 3. Mean Differences by Group on Dependent Measure: Positive Interactions Across Race Population Mean Std. Deviation Total (n=4,380).00 1.00 Asian/Asian American (n=686).28.99 African American (n=224).33.88 Latino (n=388).17 1.03 White (n=3,082) ).11.99 (A) race (B) race Mean Difference Sig. (A ) B) White Latino ).28.00 African American ).44.00 Asian/Asian American ).38.00 Note: The population mean for the entire sample was centered at zero, using a z-score transformation. Mean differences between groups are significant at p<.001 level. Only significant differences are shown. values for positive interactions between White and African American students, between White and Latino students, and between White and Asian students. Interestingly, no significant differences exist between Asian, African American, and Latino students. The lower population mean for White students coupled with the significant mean differences with the three other groups suggest that racial/ethnic boundary lines may be harder for White students to cross in the early college years. Nonetheless, a review of the regression models suggests that this may not be due completely to White students inabilities to cross those lines, but also to the availability of opportunities for interaction, especially given the varying levels of structural diversity across institutions. Indeed, as informed by our conceptual framework that highlights the role of enlightenment and mediating processes on the quality of students crossracial interactions, a closer examination of each regression model reveals important insights into the varying factors that influence each racial group. African American Students In order to adjust for the small African American sample (n = 224) in comparison to the other groups, we utilize a more lenient significance level (p < 0.10) for this group s regression model. For African American students, our model explains 24% of the variance in the dependent

20 SAENZ, NGAI, AND HURTADO variable (adjusted R 2 = 0.24) indicating a strong model specification. (See Table 4 for the results of this regression model.) Students SAT score (r = 0.14, p <.05) and their socioeconomic status (r = 0.15, p <.05) are strongly correlated with the outcome measure. Nonetheless these two measures are shown to be positively correlated with each other (r = 0.34, p <.01), and they are no longer significant predictors in the final model. Among pre-college measures in the second block, frequency of studying with individuals from different racial/ethnic backgrounds (b = 0.13, p <.05) and frequency of interactions with diverse peers (b = 0.13, p <.10) are associated with higher levels of positive interactions at the end of the second year. African American students that reported having predominantly white pre-college racial environments (b = 0.15, p <.05) are likely to have enhanced positive interactions with diverse peers in college. Of the mediating processes in the third block, only students who report strong pre-college support for diversity education (r = 0.18, p <.05) have any significant correlation with the outcome, yet this relationship is explained once the model accounts for African American students who join an organization that promotes cultural diversity. Conversely, students pre-college intergroup anxiety (b = ).13) has a negative effect on positive interactions, although its impact is only marginally significant. These findings indicate that African American students are strongly influenced by pre-college experiences and mediating processes that may already predispose them towards more quality interactions with diverse peers during college, a point also supported by previous research (Saenz, 2005; Hurtado et al., 2002). Within the college experience blocks that focus on student opportunities for enlightenment, two of the significant factors are students who report that a faculty member has taken an interest in their development (b = 0.19, p <.001) and students who report many hours per week spent socializing (b = 0.13, p <.001). For these students, what appears to be most important to positive interactions are individual social skills as well as faculty support and validation in college. Opportunities for intensive dialogue in class (b = 0.14, p <.05) also has a positive relationship to the dependent measure for African American students. In contrast, students who perceive more racial tension (b = ).16, p <.05) within the college campus are likely to report lower levels of positive interactions with diverse peers. Given the negative qualities associated with intergroup anxiety and perception of racial tension, it is reasonable to associate increased anxiety toward others and hostile campus climate with a decrease in the likelihood of positive interactions across race. Although marginally significant (p < 0.10), the finding that participation in diversity courses (b = ).12) leads to fewer positive interactions

POSITIVE INTERACTIONS ACROSS RACE 21 TABLE 4. Predictors of Positive Interactions (African American and Latino Students) African American (n=224) Latino (n=388) Variables Adj R 2 r Final beta (b) Adj R 2 r Final beta (b) Background Characteristics SAT.02.14*.00.02.16**.00** Gender.01.01.02.04.12*.17* Socioeconomic status.02.15*.02.04.08.00 Pre-college Environment Pre-college racial environment.04.21**.15*.04.12*.06 Extent of interactions with individuals from different racial/ethnic backgrounds in HS Frequency of studying with individuals from different racial/ethnic backgrounds in HS.06.19*.13 +.11.27***.18*.08.26***.13*.18.36***.14*** Pre-college Attitudes Pre-college support for diversity education.09.18*.13.24.36***.15** Pre-college level of attributional complexity.09.06.03.24.14**.04 Pre-college level of intergroup anxiety.10 ).13 ).13.24 ).11* ).09 Institutional Characteristics Structural diversity (Percentage of underrepresented minorities).10.03.00.24 ).05.00

22 SAENZ, NGAI, AND HURTADO TABLE 4. (Continued) African American (n=224) Latino (n=388) Variables Adj R 2 r Final beta (b) Adj R 2 r Final beta (b) Peer Groups Lived on campus.09.00 ).16.25.16***.01 Participated in leadership training.09.11.05.25.21*** ).06 Participated in Greek organization.09.09 ).11.26.14**.19 Joined organization(s) that promote cultural diversity.10.13.11.27.19***.05 Campus-facilitated Diversity Initiatives Participated in diversity co-curricular activities.11.14*.08.31.35***.13* Participated in diversity course(s).11.14 ).12.31.23*** ).01 Had opportunities for intensive dialogue.16.23***.14*.32.31***.11* between students of different backgrounds in classroom(s) Participated in service learning course(s).15.10 ).07.32.21***.04 Took advantage of academic support services.15.04.00.33.22***.15** Faculty demonstrated interest in development.19.20**.19***.33.21***.06 Additional Factors Perception of racial tension.20 ).09 ).16*.34.18***.09 Hours per week socializing.24.27***.13***.39.34***.16*** Significance levels: + p<0.10; *p<0.05; **p<0.01; ***p<0.001. The unstandardized beta (b) value for SAT is.00 due to a difference in scale relative to other variables in the analysis.

POSITIVE INTERACTIONS ACROSS RACE 23 is due to the fact that the measure opportunities for intensive dialogue (b = 0.14, p <.05) has a sign-reversal or suppressor effect on diversity coursework, indicating a strong inter-correlation between these two measures (r = 0.57, p <.01). The strength of this relationship suggests that diversity coursework tends to employ pedagogical techniques that encourage more opportunities for dialogue between diverse students. The suppressor effect is important because it reveals that, once we control for this tendency, those diversity courses where such dialogue opportunities are not available can actually have a negative effect on the quality of student interactions. Latino Students As a group, Latino students report significantly higher levels of positive interactions as compared to White students. The results for the Latino sample (n = 388) yield a model that accounts for the largest percentage of variance (adjusted R 2 = 0.39) in the dependent measure as compared to the other racial groups. (See Table 4 for the results of this regression model.) This suggests a strong fit for the Latino sample and indicates a good selection of explanatory and predictive measures in our regression model. Within block one of the model, which includes background characteristics, Latino students with higher SAT scores (b = 0.001, p <.01) and Latina females (b = 0.17, p <.05) are both more likely than their respective counterparts to report higher levels of positive interactions. Among the second block of pre-college measures, frequency of studying with diverse others prior to college (b = 0.14, p <.001) is one of the strongest predictors of positive interactions for Latino students, as is frequency of pre-college interactions with diverse peers (b =.18, p <.05). For the mediating process measures in the third block, students who indicated a strong pre-college support for diversity education (b = 0.15, p <.01) were more likely to report higher levels of positive interactions with diverse peers. Again, we find that pre-college experiences and selected mediating processes are most salient for Latino students in terms of impacting the quality of their interactions with diverse peers during college. Within the college experience blocks that focus on student opportunities for enlightenment, Latino students who participate in diverse co-curricular activities (b = 0.13, p <.05), who report using academic support services (b = 0.15, p <.01), and who report having opportunities to engage in intensive dialogue with students of different racial groups (b = 0.11, p <.05) are more likely to report higher levels of