Gender, Race and Ethnic Segregation

Between and Within Colleges

Jerry A. Jacobs

Department of Sociology

University of Pennsylvania

November, 1996

Draft report to the Mellon Foundation, whose support for this research project is gratefully acknowledged. An earlier draft of this report was presented at the American Sociological Association Meetings, New York, August, 1996. The research assistance of Keren Polsky, Julie Goldsmith and Erica Effler is greatly appreciated.

Abstract

In this study I examine the extent to which women and minorities are unevenly distributed across colleges and universities in the United States. I also analyze the extent of segregation by majors by sex and race within colleges. I employ data on 1,165,582 bachelor's degree recipients in 1,554 institutions of higher education in 1993 from the HEGIS/ CASPAR data base.

The stratification of women into lower-status schools than men is modest. Only 14.5 percent of women would have to change schools to be distributed in the same manner as men. Women trail men slightly in representation in high status institutions, as measured by average SAT score, acceptance rate, percent of faculty with PhDs, student/faculty ratio, first year retention rate, graduation rate, and in-state tuition fees. The sex stratification between schools that does exist can be explained by two factors: 1) the greater representation of men in schools with engineering programs; and 2) by the greater representation of women among part-time students, who are less likely to be found in high status institutions. Women, however, remain segregated from men in the fields of study pursued. Nearly 30 percent of women would have to change majors to be distributed in the same manner as men. This report examines factors that promote or inhibit the gender segregation of majors within schools.

Race and ethnic segregation between schools is much higher than gender segregation. This study documents substantial social isolation of minority students. A large majority of white students obtain their degrees from institutions with limited social diversity. The high level of segregation between African Americans and whites is partly due to the continued enrollments of large numbers of African Americans in Historically Black institutions. Asian-American and Hispanic degree recipients are even more segregated by institution from their white counterparts than are African Americans. For these groups, geographic representation in a limited number of states contributes to the high level of segregation between institutions. African-American and Hispanic students obtain their degrees in less prestigious institutions than do whites, while Asian Americans are found in somewhat more prestigious institutions.

The segregation of majors by race and ethnicity follows a more complicated pattern. Nationally, there is much less segregation of majors by race and ethnicity than is the case by gender. However, within institutions, minority groups and whites are highly segregated by majors. I speculate that minority groups cluster in particular majors in predominantly white institutions as a form of social support in what can be an unwelcoming environment. In the final section of the report I examine the factors that promote or inhibit gender, race and ethnic segregation on college campuses.

In the conclusion I discuss the significance of these findings for our understanding of race, ethnic and gender stratification in American society. I maintain that studies of higher education need to examine the college experience, and not just differential access to higher education. The differences highlighted by sex, race and ethnicity are related to the educational history of these groups, and to their position in the occupational structure.

When I began this project, I intended to focus my research on differentiation by field of study within colleges. The goal was to build on my earlier work on the gender segregation of college majors (Jacobs, 1995) by analyzing variation between schools, and by adding race and ethnicity to the analysis. I immediately had to decide whether to include single sex schools, and those that were single sex (or single race) de facto if not de jure (i.e., those that graduated only men or only women in a given year). I also had to decide whether to include schools where there was no segregation among students -- where all students received degrees in the same field of study. These details forced me to distinguish between two related issues -- the extent of segregation (by sex, race and ethnicity) between schools, and the extent of segregation within schools. One can easily imagine an educational system in which women pursue art and music at one set of institutions and men pursue engineering and science at another set of schools. In a world dominated by single-purpose institutions, the segregation of majors and schools would coincide. We generally think of our system of higher education as approximating a university model where most if not all subjects are available at most schools. In such a model, differentiation between men and women in subject matter would typically be evident within individual schools. I realized that it would be useful to establish how much differentiation occurred within schools and how much between schools.

The first section of this report examines the extent of gender, race, and ethnic differentiation between institutions of higher education. I ask what proportion of women (and African Americans, Asian Americans and Hispanics) would have to change schools in order to obtain a proportional distribution. A related question is whether gender, race, and ethnicity are associated with the stratification of colleges and universities. In other words, are men and women largely attending schools of the same caliber? I examine the factors that contribute to the gender, race and ethnic differentials in school standing. This analysis highlights the distinctive patterns of gender, race and ethnic segregation in American higher education. The second section of this report is devoted to the issue of segregation by major within schools.

1. The Distribution Across Institutions by Gender, Race and Ethnicity

Men's And Women's Representation Among Colleges

I began this research with the premise that gender inequality is overwhelming an intra-school phenomenon. I assumed that men and women are equally represented in low-ranking and high-ranking schools. I expected that only a very small proportion of women would have to change schools in order to be distributed in the same manner as men. Women's colleges have received substantial attention from researchers, but students at these schools constituted only 1.3 percent of women degree recipients in 1993. Very few schools are restricted to men, despite the significant attention given to court challenges to the practices at the Citadel and the Virginia Military Institute.

On further investigation, I found several studies that suggested that women are not equally represented at top-tier institutions. Hearn (1990) and Persell et al. (1992) reported that women were disadvantaged in access to elite schools, based on an analysis of data on 1980 high school seniors. Hearn found that 45 percent of students at top institutions were female, compared with 40 percent in moderately selective and 51 percent in non-selective colleges, based on his analysis of the High School and Beyond data. (Hearn defined elite schools as those with average SATs above 1,176 in 1980. Persell et al. defined selective schools as those where entering students had an average SAT of 1050 or more.) The average young woman graduating from a public high school was estimated to have a 4.2 percent chance of attending an elite college, compared with 5.8 percent for male students. Persell et al. also examined the college attendance patterns of graduates of elite boarding schools. Just as high a fraction of female as male graduates of these schools attended elite schools, but being female had a negative effect once other factors were controlled.

On the other hand, women have made progress since 1980, both in overall enrollments in higher education as well as in access to elite institutions (Karen, 1991). Karen reports that the proportion of women at Ivy League schools rose from 22.5 percent in 1960 to 43.3 percent in 1986, and at other prestigious institutions (excluding the Seven Sister Schools) it increased from 26.4 percent in 1960 to 46.7 percent in 1986.

Reports on admissions and enrollments from selected leading institutions indicate that women are well represented among recent entering classes (Monthly Forum on Women in Higher Education, 1995). Table 1 reproduces the Monthly Forum data on leading universities and leading liberal arts colleges. One conclusion that is immediately obvious from the university figures is that schools that prominently feature engineering programs often have substantial majorities of male students. The three institutions with less than 40 percent women among entering freshmen were California Institute of Technology, Johns Hopkins and Carnegie Mellon, all schools with large engineering programs. On the other hand, of the 24 leading schools shown in Table 1, 13 had more than 50 percent women among entering students.

The representation of women among the top-tier liberal arts institutions was even more favorable. The lowest percentage female of the schools included in this list were Washington and Lee, with 41.9 percent, Claremont, with 44.9 percent and Amherst, with 45.7 percent. Most of the others had a female majority, with Oberlin, at 60.2 percent female, enrolling the largest percentage of women. Of the 20 schools listed (excluding the three all-female schools), 14 enrolled more than 50 percent women in the 1995 freshman class. It is hard to discern systematic under-representations of women at these institutions.

HEGIS / CASPAR Data

I examined the distribution of men and women among colleges with data on earned degrees conferred. The data were assembled by the Department of Education in the HEGIS data base.⁽¹⁾ They have been made accessible to researchers on line via the CASPAR system.⁽²⁾ These data have many virtues: they represent a comprehensive account of all degree recipients, they are available annually during the period 1966-1993, and they allow for analysis between schools as well as variation by major within schools. The analysis presented here develops a clear picture of the relationship between gender, race and ethnicity and bachelor's degree attainment. (Of course, these results would be misleading with respect to college entrance if the groups' completion rates differed sharply.)

Table 2 summarizes the number of schools included in the study, and details the reasons various groups of institutions were excluded. I began the selection process using the 3905 academic institutions included in the HEGIS/ CASPAR database for 1993. I then excluded 1332 two-year institutions, based on the Carnegie classification system.⁽³⁾ Of the 2573 remaining schools, 850 granted no bachelor's degrees in 1992, leaving 1723 schools. The between-schools analyses presented below are based on these 1723 colleges and universities. Excluding single sex schools brought our total down to 1609. (Single sex schools are included in the between-school analyses, but not the within-school analysis.) Lastly, for certain analyses I excluded the 55 institutions that granted fewer than 20 bachelor's degrees in 1993, leaving a total of 1554 institutions that include 98.95 percent of graduates. Included in this set of schools are many separate campuses of large universities. I felt that each campus should represent a single case, because the campus is where students meet other students and where their fields of study are chosen.

Dimensions of Segregation

There are at least three distinct aspects of segregation (Jacobs, 1993; Massey and Denton, 1989). I will illustrate with examples from gender segregation because they are the most familiar to me. The principal dimension of segregation that is the focus of most research is the degree to which groups are distributed unevenly across fields. This concept is typically measured with the index of dissimilarity (D), which indicates the proportion of one group that would have to change fields to be distributed in the same manner as another group. The same statistic can be used to measure segregation between institutions. Recent studies of the U. S. economy indicate that over half of the employed women would have to change occupations and over two-thirds would have to change job titles to be distributed in the same manner as employed men (Jacobsen, 1994; Tomaskovic-Devey, 1993). The level of labor force sex segregation declined during the 1970s and 1980s after remaining largely unchanged for most of the century (Jacobs, 1989). Similarly high levels are observed in other industrial countries as well as developing countries (Charles, 1992; Jacobs and Lim, 1992). As we will see below, just under one third of women undergraduates would have to change fields to be distributed in the same manner as their male counterparts.

I also employed another measure of segregation, Theil's H, which adjusts for random variation that occurs when a small minority group is distributed across a large number of units (Theil and Finizza, 1971; Theil, 1972; Massey and Denton, 1988; White, 1986). Whereas the index of dissimilarity compares the observed distribution to one of perfectly proportional representation, H compares the observed distribution with a random distribution. In cases such as U. S. colleges and universities, where many schools have very small numbers of minority students, the adjustment for random variation may make a notable difference. Unfortunately, H does not have the easy interpretability of the index of dissimilarity.

A second feature of segregation is the crowding of one group into a limited number of fields (or crowding into a limited number of institutions). This aspect is not directly captured by the index of dissimilarity, and requires the use of specific indices of concentration, designated RC, for relative concentration or crowding.

Crowding is important for two reasons. First, crowding is an indication of the extent of opportunities for women (or minority groups). While all degree fields are now formally open to both men and women, some fields, such as engineering, remain de facto male preserves. The concentration of a great majority of women into a handful of fields would be one indication of how pervasive social restrictions are on women. For example, in 1960 almost half of women receiving bachelor's degrees did so in one field, namely teaching. Over 75 percent of women received their degrees in one of six fields: English, fine arts, history, home economics, nursing and teaching. This situation had changed dramatically by the 1980s (Jacobs, 1995). Similarly, differential concentration between institutions may reflect more limited opportunities for one race or ethnic group compared with another, although other factors, such as geographic concentration, may contribute to the over-representation of some groups in particular colleges and universities.

A second reason to examine the dispersion of men and women across fields is that the financial potential of a field is influenced by the relationship between supply and demand. Edgeworth (1922) argued that women earned less than men in part because they were crowded into a limited number of fields. (The issue of crowding is also discussed by Bergmann, 1986, and Parcel, 1989.) Restricting women to a narrow set of jobs approved of as "women's work" produced an excess supply of women for these occupations, thus limiting women's bargaining power and lowering their wages. The extent to which women are crowded into a few fields of study is one indication of the potential economic returns to their degrees.

A third aspect of segregation is the degree of intergroup contact, in other words, the chances of men and women sharing an institution or a field of study. The intergroup contact index -- designated P* by Lieberson (1980) -- reflects both the level of segregation and the representation of each group. While the index of dissimilarity is symmetrical -- it is the same from the point of view of the majority or the minority group -- measures of inter-group contact can be asymmetrical. In other words, women's chances of sharing a field of study with men differ from men's chances of sharing a field with women. One striking result of the growth in women's labor force participation is that women's chances of sharing an occupation with men have declined (as working women's numbers increase, women's chances of sharing an occupation with another woman increase) while men's chances of sharing an occupation with women have increased markedly. The two groups thus differ in how they experienced the same changes, an aspect of segregation revealed by measures of inter-group contact. The same pattern is evident in the segregation of college majors, as the number of women enrolled in college continues to grow.

Contact measures are also informative for our analysis of differentiation between institutions. Contact between white and minority students will be heavily influenced by the distribution of these groups across institutions. Contact indicators will assess the extent to which white students are likely to encounter minority students on their campus. Segregation within institution by major may compound the uneven distribution of groups between schools.

I will begin with the issue of differentiation between institutions, and then turn to questions regarding the distribution of majors within institutions. Initially I will present information on the extent of segregation between schools by gender, race and ethnicity. I will then explore the extent to which this segregation represents differential representation at high status institutions.

Sex Segregation Between Schools

Women comprise the majority (54.4 percent) of bachelor's degree recipients in 1993. Women are no longer a disadvantaged minority in terms of access to higher education (Jacobs, 1996). However, differentiation between men and women may exist between institutions, or by major within institutions. Table 3 presents a variety of complementary measures that indicate the extent of segregation between men and women in the colleges and universities where they earn their degrees.

Overall, in 1993, 14.6 percent of women would have had to change schools in order to be distributed in the same manner as men. While this figure is somewhat higher than I expected, it is low relative to most measures of segregation. For example, segregation by sex by major within schools is typically about 30, more than twice as high as segregation across schools. As we will see, the level of segregation between schools by race and ethnicity is much higher.

Another approach to the question of segregation is to tap the degree to which one group is crowded into a limited number of institutions. In 1993, men were slightly more crowded into a limited set of schools than were women (RC= -3.7). This reflects the fact that men are more likely to enroll in very large institutions. The top 100 schools graduated 36.4 percent of men, compared with 32.4 percent of women. (As we will see, racial minorities are much more crowded into a limited set of institutions than are their white counterparts.)

Single sex institutions play a minor role in the institutional segregation between men and women. There were 72 schools with only male graduates, nearly all schools offering religious instruction, with VMI, Citadel and a few engineering and technical schools comprising the balance. Only 3,375 men received bachelor's degrees at these schools, 0.6 percent of male degree recipients. Another 42 schools graduated only women, which included 8,437 graduates (which represented 1.3 percent of women bachelor's degree recipients that year). Single sex schools by themselves are responsible for very little segregation by sex between institutions.

A third way to approach the question of segregation is to ask what share women represent in the average class in which women graduate. This is a measure of group contact that reflects both the level of segregation and the overall gender makeup of U. S. colleges and universities. In 1993, the typical woman graduated in a class that was 55.6 percent female, while the typical male received his diploma in a class that was 53.0 percent female. Overall, the extent of gender segregation between institutions is modest. As we will see in the next section, institutional differentiation by race and ethnicity is far sharper.

Racial and Ethnic Differentiation between Institutions

1. African Americans and Whites

African Americans remain under-represented in higher education. Approximately one quarter (24.5 percent) of 18-24 year old African Americans were enrolled in college in 1993, compared with 36.8 percent of whites (figures obtained from the National Center for Educational Statistics, 1995). Of high school graduates, just over one-third (35.6 percent) of African Americans aged 18-24 were enrolled in college in 1993, compared with 43.7 percent among whites. These participation rates represent progress compared with 20 years earlier, yet, because enrollment rates continued to increase among whites as well, African Americans were unable to close the gap.

The increasing enrollment of African Americans in college and universities surely is a positive development for their economic prospects, because educational credentials facilitate access to rewarding jobs. However, the economic benefits students derive from higher education in part reflect the schools they attend and the fields of study they select.

In this section I focus on the extent of segregation between African Americans and whites between schools of higher education. We have seen that the disparity between men and women is modest. How does the gender gap in access to high status schools compare to the race differential? Results for 1993 are presented in Table 4.

Overall, 46.3 percent of African-American students would have had to change schools in order to be distributed in the same manner as white degree recipients. Theil's H is also substantial (27.2), which indicates that the segregation of African-American from white students is not a matter of random variability. Just under half (45.1 percent) of African Americans obtained their degrees from the most popular 100 institutions, compared with just over one-third (34.7 percent) of whites. The relative concentration statistic also indicates that African Americans are more crowded into a limited set of schools than are whites (RC=10.1). One underlying reason for this discrepancy is the continuing role of Historically Black Institutions in the higher education of African-American students. In 1993, 37.8 percent of African-American students obtained their degrees from Historically Black Institutions, as did only 0.6 percent of white students. In 1993, 27 institutions graduated only African-American students. At the other end of the spectrum, 1066 schools graduated fewer than 20 African-American students, including 218 schools with no African-American graduates. I found this pattern quite dramatic. At most colleges, whites graduate with only a small number of African-American classmates.

This fact may well have important consequences for both African-American and white students. At majority-white institutions, African Americans may be offered a limited set of same-race friends, and yet may feel the need to stick close together in the face of an over-whelming majority of white classmates. Whites may have limited contact with African- American classmates, and may not have the opportunity to get acquainted with them as individuals.

These dramatic disparities in representation result in sharply different patterns of contact between African-American and white students. The average female African-American degree recipient graduated from a school with 36.1 percent African-American students, with nearly one third (32.2 percent) obtaining their degrees at schools in which African-American students comprised the majority of degree recipients. In schools other than Historically Black institutions, the share of their fellow graduates who are African American is 15.6 percent, with 19.9 of those in the same major being other African-American women. In contrast, the average white graduates from a school with less than 4.5 percent minority degree recipients, and only less than one percent (0.3 percent) obtained their degrees from institutions in which African-American students represent the majority of graduates. Over three in five (62.2 percent) white students obtained their degree in a school with less than five percent African-American graduates.

The results presented here resemble in broad outline data made available by the National Center for Educational Statistics (1995, Table 202, p. 210). There it is revealed that the majority of U. S. students attend schools with fewer than 20 percent minority students, while nearly half of minority students are enrolled in schools with over 40 percent minority students. My analysis further differentiates among the college population, and highlights the different experiences of white, African American, Asian and Hispanic students.

These results are quite revealing of the continued racial segregation in higher education in the United States. Discussions of diversity in colleges and universities must begin with the fact that our system of higher education remains sharply divided by race. The majority of white college graduates receive their diplomas with few if any minority students joining in the procession. As we will see, the situation of African-American students is not unique. Other minorities, such as Asian Americans and Hispanics, are also highly unevenly distributed across colleges and universities.

2. Asian Americans

The HEGIS/ CASPAR data include information on Asian-American students, but they do not allow for further differentiation by ancestry among Asian Americans. There is much diversity among Asian Americans, for under this umbrella are included individuals of Japanese, Chinese, Filipino, South-East Asian, Pacific Islander and other ancestries. There is reason to expect diversity among these groups, based on the recency of immigration, parental socio-economic resources, and other factors. Nonetheless, in the analysis we examine the degrees awarded to Asians without further differentiation. This is because of the data limitations already mentioned, but also because further differentiation would be difficult to assess empirically. Asians constitute a small fraction of degree recipients, representing fewer than 5 percent of bachelor's degree recipients in 1993. There are many schools in which there are no Asian students, and large numbers of schools in which these groups represent only a handful of degree recipients. There are limits on how small groups can become before the analysis stretches the usefulness of the data.

Asian Americans exceed the national average in college completion. Over 40 percent of Asian and Pacific Islanders in the U. S. in 1994 had completed their bachelor's degree, compared with under 25 percent of the entire population (U. S. Bureau of the Census, 1996). Exploring the college experience for Asian Americans enables us to determine whether their distribution across schools and majors reinforces their advantaged position or mitigates it. We will also explore issues of contact and isolation from other groups of college students.

Asian-American students are far from evenly distributed across colleges and universities in the U. S. Overall, just over half (52.0 percent) of Asian-American students would have to change schools to be distributed in the same manner as whites. Theil's H is lower for Asian-Americans than for African-Americans, which suggests that the higher index of dissimilarity may be due in part to random variability across schools for Asian Americans, who constitute a smaller minority of college students than do African Americans. Nearly two-thirds (65.7) of Asian-American students receive degrees from the most popular 100 institutions, compared to just over one third (34.7 percent) of white students. There were only eight schools that graduated only Asian-American students. (On closer inspection, these are programs with just a few graduates and do not really represent exclusively Asian institutions.) At the other end of the continuum, 1334 institutions graduated fewer than 20 Asian-American students, with 368 schools graduating no Asian-American students.

This highly uneven distribution of Asian-American and white students results in sharply skewed contact between these groups in the college setting. The average Asian student graduates from a college with just under one-fifth Asians (19.0); the average white has only 4.4 percent of the graduating class comprised of Asians. The great majority of Asian-American students graduate from schools with large majorities of whites. Less than one in ten (6.8 percent) obtains a degree from a school in which Asians represent the majority of students.

One quarter (26.5 percent) obtain degrees in schools where Asian-American students represent just five percent of the student body. The typical white student graduates from an institution with even fewer Asian-American students. More than three of four (78.5 percent) Asian-American students graduate from a school with less than five percent Asian Americans in his or her class.

The segregation of Asian students by school reflects the fact that they are concentrated in just a few states. Nearly three in five (58.6 percent) Asian-American students receive their degrees in one of five states: California (33.4 percent); New York (10.4 percent); Hawaii (5.0 percent); Illinois (5.0 percent) and Texas (4.8 percent). In contrast, under one third of whites (31.6) and African-American (31.6 percent) degree recipients are located in the five most populous states.

3. Hispanics

Hispanics are nearly as diverse as Asian Americans. Included among Hispanics are individuals with ancestry in Mexico, Puerto Rico, Cuba and other Caribbean and Central American countries. In this analysis we are not able to explore this diversity. (A regional analysis might succeed in separating individuals with Mexican and Central American heritage from others of Hispanic origin.)

Hispanic students are under-represented among college students. Of those aged 18-24, 21.7 percent of Hispanics were enrolled in college in 1993, compared with 36.8 percent for non-hispanic whites. This low figure is principally due to the low high-school graduate rate among Hispanics. The Hispanic-Anglo gap is smaller among high schools graduates (36.1 percent for Hispanics, versus 43.7 percent for Anglos).

In addition to this marked disparity in educational attainment, Hispanic students have a sharply different college experience than do white students. Hispanic students are the most segregated between schools of any group studied here. Overall, 57.5 percent of Hispanic students would have to switch colleges and universities in order to be distributed in the same manner as whites. An even more dramatic result is that over four of five Hispanic students (82.7 percent) are found in the top 100 institutions, compared to just over one-third of white students. There were 1324 schools in which fewer than 20 Hispanic students obtained their degrees, including 347 in which there were no Hispanic students.

Principally because of a number of campuses located in Puerto Rico, one third (33.9 percent) of Hispanic students graduate from schools in which they represent the majority of students. Nonetheless, Hispanic students typically find themselves in the minority. The average Hispanic degree recipient is part of a class with just over one third (37.6 percent) Hispanic students. In contrast, the average non-hispanic white student has less than one out of 20 (3.9 percent) Hispanics in his or her graduating class. The vast majority of non-Hispanic white students (76.5 percent) graduate from schools with less than five percent Hispanic graduates. Only 0.4 percent of non-Hispanic whites graduate from schools in which Hispanic graduates represent the majority.

The concentration of Hispanic students reflects the geographic distribution of the Hispanic population. Over 70 percent (71.8 percent) of Hispanic students receive their degrees in four states and the island of Puerto Rico: Puerto Rico (23.1 percent); California (18.4 percent); Texas (13.9 percent); New York (9.0 percent) and Florida (7.4 percent). Given this geographic concentration, a high degree of segregation by institution is inevitable. However, even if we were to leave the case of Puerto Rico aside, Hispanic students were still highly unevenly distributed across schools. The index of segregation for the U. S. excluding Puerto Rico is 51.5, down from 57.5 when Puerto Rico is included. Theil's H is 22.2 when Puerto Rico is excluded, which is about the same level as that observed between Asian Americans and whites.

Diversity from the Vantage Point of White Students

There are large numbers of schools with few African Americans, Asian Americans, and Hispanics. The analysis presented above examines contact between whites and minority groups one at a time. Another way to think about diversity from the point of view of white students is to consider the likelihood of graduating with students from any minority group.

I conducted an additional analysis that included all minority students, including Native Americans, and foreign students. I asked what fraction of any of these groups are represented in the average school attended by non-Hispanic white students. Nationally, minorities and foreign students represent 19.9 percent of bachelor's degree recipients. But that does not mean the typical white student graduates in a class with 19.9 foreign or minority students. The average non-Hispanic white graduate in 1993 obtained his or her degree in a class that was 85.5 percent U.S. born white, or 14.5 percent minority or foreign. Half of white students obtained their degrees from schools that were 88.8 percent or more U. S. born white; 42.5 percent of whites obtained their degrees from schools that were at least 90 percent white and U. S. born, while 79.2 percent obtained their degrees in schools with at least 80 percent white U. S. born graduates. Only 1.7 percent of white students obtained their degrees from schools in which U. S. born whites were a minority of graduates. Adding together all the diverse minority groups and foreign students results in a picture of whites having more contact with a diverse body of students. In other words, while large majorities of white graduates have only limited contact with individual minority groups, taken as a whole contact with minority group and foreign students is more common, but far less than would be expected by chance.

Summary

The results indicate that gender, race and ethnicity vary in their effect on the distribution of students across colleges and universities in the United States. Gender differentiation between schools is rather modest, as the typical school from which women graduate has only a few percentage points more women than does the typical school from which men graduate.

Race and ethnic differentiation is much more marked. There are literally hundreds of colleges and universities with no African-American, Asian-American or Hispanic graduates. The majority of campuses in this country graduated fewer than 20 of each of these minorities in 1993. Minority groups are unevenly distributed for different reasons. African-American graduates continue to be concentrated in Historically Black institutions. Asian-American and Hispanic students' representation is highly skewed by geography.

These data indicate a sharp pattern of differentiation between institutions based on race and ethnicity that far exceeds the institutional segregation by sex. The results indicate that many institutions of higher education in the United States are limited in their social diversity. White students often obtain their degrees in schools that are overwhelmingly white. Whatever the causes of this pattern, the high degree of racial and ethnic segregation between colleges and universities is relevant to the ongoing discussions of cultural and social diversity that continue to preoccupy much of the debate surrounding American higher education. Coeducation is the norm in higher education in the United States, while racial and ethnic diversity is all too often the exception to the rule.

Inequality in School Standing

The analyses just presented assess the degree of differentiation between schools but do not consider how schools rank in terms of prestige or quality. Were women and minorities concentrated in low status schools? I merged a variety of data on college rankings with the HEGIS/ CASPAR data on degrees earned. These data are derived either from HEGIS information on college standing or from U. S. News and World Report data on colleges. The measures I employed included: average SAT, acceptance rate, percent of faculty with PhDs, student/faculty ratio, graduation rate, first year retention rate, in-state tuition fees, and percent enrolled part time.⁽⁴⁾ While one might raise questions regarding the validity of each of these measures, taken together they are surely related to the attractiveness, selectiveness, status, or "quality" of colleges. As we will see, the results indicate strong commonalities in the behavior of these measures.

Table 5 presents means of these indicators, weighted by the distribution of male and female degree recipients. In 1993, women remained slightly over-represented in lower status institutions than their male counterparts. The differences between men and women were not great, but they were consistent across a wide range of measures that are indicative of institutional status or quality. Women, compared to men, received degrees from institutions with slightly higher acceptance rates, lower standardized test scores, lower fees and more part-time students.

The sex gap in college status documented in Table 5 can be explained by two factors: 1) the relative scarcity of women in schools with large engineering programs, because engineering programs are typically male-dominated and tend to be more selective than most other programs; and 2) the tendency of women to enroll in school part-time, because lower-status institutions are more likely to accept part-time students.

Table 6 presents several regression equations which demonstrate that these two factors account for the sex gap in college quality. I present several nested comparisons. In the first equation, the percent female in a school is the sole independent variable. In the second equation, I add the percent of students receiving their degrees in engineering. In the third equation, I add the percent of students enrolled part time. Average SAT score is the dependent variable. The first equation shows that women trail men in this measure of college standing. Specifically, the greater the proportion female, the lower the average SAT scores of a school. However, the effect is cut by more than half when the proportion of engineering majors is controlled. When the proportion of part time students is added to the analysis, the SAT gap declines even further, and the differential is no longer statistically significant.⁽⁵⁾

Engineering is not the only subject that contributes to the sex stratification of schools. Education tends to have the same effect for the opposite reason. Education programs tend to enroll students with below-average SAT scores, and women are over-represented in such majors. If I substituted education for engineering in Models 2 and 3 in Table 6, the results would be similar to those presented. The gender coefficient would fail to be statistically significant in the third model, but the education coefficient would be negative. Business students have below-average SATs, but because business is now a relatively gender-integrated field of study, the size of a business program does not significantly affect the gender stratification of colleges and universities.

In the second panel in Table 6, similar findings for acceptance rates are presented. In this case, schools with a larger share of women have higher acceptance rates. Engineering enrollment has relatively little effect on this pattern, while part-time enrollment is the principal factor that accounts for the sex differential. The analysis of graduation rates shown in the third panel of Table 6 follows the same pattern. The sex differential disappears once engineering and part-time enrollment are partialled out of the analysis. For each of these measures, as was the case for SAT scores, the sex differential is no longer statistically significant after these two factors are accounted for.

I was especially interested in the SAT differential, since earlier studies relying on SAT scores found women under-represented in elite schools. The analysis of the sex gap in SAT scores presented in Table 6 showed that sex difference in school means could be accounted for by the two variables examined, percent enrolled in engineering and percent enrolled part time.

However, this similarity in means could be consistent with women's under-representation at the high end of the distribution, namely elite schools. I sought to more closely replicate the research of Hearn and Persell by using cutoff scores for elite schools. I used several measures in order to determine whether the results were sensitive to particular cutoff values. Thus, I employed three definitions of elite schools: those with an average SAT score of 1200 or better, 1100 or better and 1000 or better.

I conducted regression analyses which parallel those presented in Table 6. However, the dependent variable now is a dichotomous measure. Each school is scored 1 if it is above the cutoff value, and 0 otherwise. As a result, it is appropriate to use logistic regression analysis rather than ordinary least squares regression. Instead of presenting R2 as a measure of the explanatory power of these equations, I switch to the proportion reduction in chi-squared (I used L2, or the log-likelihood measure of chi-squared as my indicator.) These results are presented in Table 7.

The substantive result is the same here as was evident in previous analyses: once the percent of students receiving degrees in engineering and the percent of students enrolled part time are controlled, the relationship between sex composition and elite status is severely attenuated and fails the conventional test of statistical significance. This finding is robust for the three cutoff values employed. ⁽⁶⁾

Programs such as engineering and education significantly alter the distribution of men and women across schools in terms of selectivity. They serve to segregate men and women within schools and also inhibit access of women to competitive institutions. The fact that these fields work to segregate men and women both within and between schools highlights the usefulness of examining both of these aspects of gender differentiation.

African-American Graduates

The race gap in college ranking is much more dramatic than the modest gender gap we have observed (see Table 8). The race difference in the college's average SAT score is nearly 60 points, compared to about 20 between men and women. This differential is probably understated because of missing data on institutions' SAT scores.⁽⁷⁾

The race gap in percent of faculty with PhDs, student/faculty ratio, and tuition are all substantially larger than the corresponding gender gap. The lower graduation rates and freshman retention rates at schools with larger representations of African-American students are clearly of concern. There is one notable exception to the general pattern. Acceptance rates

are actually lower in schools attended by blacks than in other schools. If lower acceptance rates are taken as an indicator of school quality, then on this measure blacks are slightly ahead of whites. On the other hand, there may be different application patterns that account for this pattern, specifically the popularity of Historically Black schools among African-American college applicants. This exception notwithstanding, the pattern of black disadvantage relative to whites in school ranking is quite consistent.

Given the prominence of Historically Black institutions in the education of African-American students, I sought to determine whether the differentials in institutional status documented in Table 8 are due to the concentration of African-American students in Historically Black Institutions.

The analyses reported in Table 9 address this question. The results are somewhat less consistent across measures for the analysis of race differences than was evident in the sex-gap analysis reported above. The SAT differential disappears when a control for the representation of African-American students in Historically Black Institutions is included in the analysis, along with controls for engineering majors and part-time enrollment. On the other hand, African Americans attend schools with somewhat lower acceptance rates, and this pattern intensifies once controls are included in the analysis. However, African Americans attend schools with lower graduation rates than do whites, and this pattern does not change substantially when controls are included in the analysis.

In analyses not shown here, I analyzed the African American versus white differential in terms of access to high-status institutions. I examined whether under-representation of African Americans persists after the control variables were included in the model. In each of these analyses, the gap could be attributed to a combination of the presence of Historically Black institutions, engineering majors, and part-time enrollment.⁽⁸⁾

Asian American Graduates

Asian students receive degrees from schools with higher average SAT scores than whites (see Table 10). All measures of school standing favored Asians except the graduation rate and instate fees, where the differences were not statistically significant. The Asian advantage in SAT scores is reduced but not entirely explained by the over-representation of Asian Americans in engineering programs (see Table 11). An Asian-American advantage persists with most measures even when the percentage in engineering and the percentage part-time are controlled.

Hispanic Graduates

Hispanic graduates trailed whites in the average SAT score of the school they attended, but some of these measures were not statistically significant (see Table 12). As was the case for African Americans, these results probably understate the differential due to missing data (see footnote 6). Hispanic students actually were found in more favorable institutions than whites on several other key measures: they attended schools with lower acceptance rates, higher freshman retention rates, higher percentages of faculty with PhD degrees, and higher in-state tuition fees. On closer inspection, however, it becomes clear that these advantages reflect the regional distribution of Hispanic students. Colleges and universities in the Northeast have higher average SAT scores than do schools in other regions. Many other school attributes, including in-state tuition fees, are also higher in the Northeast. When region is controlled, Hispanics trail whites in the school's average SAT score, first year retention rate, graduation rate, and in-state tuition fees.

Hispanics are no more or less likely to major in engineering than are whites. Consequently, engineering enrollments explain none of the Hispanic-white differential in school standing. However, Hispanic students are more likely than whites to enroll part-time, and the extent of part-time enrollment does account for some but not all of the differential between whites and Hispanics.

The regression analyses of school standing for Hispanics are presented in Table 13. As noted above, adding controls for region increases the Hispanic/Anglo differential. In other words, while Hispanics do not trail whites nationally, they do trail whites within the same region of the country. The addition of engineering majors and part-time enrollments narrows the differential, but this is due exclusively to the contribution of part-time enrollments.⁽⁹⁾

Summary

To summarize the between-school analyses presented thus far, the degree of gender stratification between institutions of higher education is relatively modest, and is accounted for by the distribution of engineering programs and by women's over-representation among students enrolled part-time. African Americans trail whites in school standing because of the high fraction of students that continue to enroll in Historically Black Institutions. The race gap in school status is partly due to the fact that these Historically Black Institutions are not among the nation's elite schools. For some measures, such as average SAT score, the race differential is accounted for by the three factors included in the analysis. For other measures, such as graduation rate, a substantial differential remains.

Asian Americans are even more segregated from their white counterparts by institution than are African Americans. However, this segregation is more a matter of geography than of school standing. Indeed, Asian Americans receive their degrees from schools with slightly higher than average measures of school standing, such as test scores, faculty credentials, and student faculty ratio. These differences are not simply a matter of the disproportionate enrollment of Asian Americans in engineering: they persist even after this factor was controlled in the analysis.

Hispanic students are the most segregated by institution of the groups examined here, undoubtedly due in part to their geographic distribution. Hispanic students are also over-represented at schools with lower test scores and other measures of school quality, once their regional distribution was taken into account. This difference was only partly explained by the over-representation of Hispanics among the ranks of part-time students.

Segregation of Majors

Single-Focus Schools

When I began to conduct my analysis of segregation within schools, I noticed a group of schools where the index of dissimilarity was zero. In other words, these were schools in which there was complete congruence in the majors chosen by men and women. Closer examination revealed that in these schools all students were listed as majoring in a single subject. I decided to call these single-focus programs.

Of the 1554 schools that I focused on in my analysis of within school segregation, there were 106 with no segregation by sex among the graduates. These break down as follows: in 31 schools all students were listed as majoring in an unclassified subject. I view these as instances of reporting failures by the administrations of these instititutions. The remaining 75 fit the description of single-focus programs. In 26 schools all students received their degrees in art; 13 graduated only nursing students, and in 10 students received degrees only in theology. The remaining 26 were other single-purpose schools, including less than 4 each for business, chiropractic studies, optometry, pharmacy, architecture, engineering, and other miscellaneous technical institutes. These 106 schools were small and represented only 1.49 percent of male degree recipients and 1.31 percent of female degree recipients in 1993. I included them in my analysis of the relative standing of men and women across universities, but I excluded them from the analysis of segregation of majors, unless otherwise indicated. Segregation within schools thus is confined to schools with both men and women students, and where graduates are distributed across more than one field of study.

Segregation by Sex Within Schools

Table 14 lists the 29 fields of study examined in this research, and shows the proportion of men and women in each field, along with the percent female of the field, based on national totals of degree recipients. A number of prominent fields can be labelled gender neutral, in that the proportion of women graduating in the field is similar to the proportion of men obtaining degrees. Biological sciences and business, while not being perfectly gender balanced, nonetheless approximate this ideal, as do unspecified social science fields. Nonetheless, a number of fields continue to be skewed to either men or women. Engineering is the most male-dominated of the large fields of study, with just under one in six (15.9 percent) of the degree recipients being women. Other life sciences (mostly nursing), education, and psychology are among the largest female-dominated fields.

There is segregation by sex within each of the specialties listed in Table 14. Among engineering graduates, women gravitate to chemical and biological engineering more than mechanical or aerospace engineering. Among educational specialties, women are more likely to pursue early childhood education than to specialize in high school math and science. Nonetheless, the 29 fields examined here capture the lion's share of the segregation that would be tapped with more detailed measures. In a related analysis, I found that 24 specialties similar to those examined here captured 88 percent of the segregation revealed with data on over 442 specialties (Jacobs, 1995).

Table 15 presents data on the level of sex segregation by field of study within colleges and universities. The first row represents the national total. This facilitates the comparison of the national picture to that within schools. Nationally, 28.4 percent of women would have to change majors to be distributed in the same manner as men. This level represents a substantial reduction from the situation in 1960, when the index of segregation was over 50. However, the level of segregation between men and women has not changed over the last 10 years (Jacobs, 1995).

How does this national picture compare to that observed within schools? Within the average college or university, 31.2 percent of women would have had to change majors to be distributed in the same manner as men, which is slightly higher than the national average of 28.4. This indicates that within-school segregation is the principal form of sex segregation in the U. S. In other words, if the segregation of majors took the form of men attending engineering schools and women attending teacher's colleges, then the level of segregation within schools would have been lower than that observed nationally. Instead, we find that the sex segregation between majors observed nationally is due to that taking place within the average college.

Women are slightly less crowded by discipline than are men. Men's concentration in business and engineering results in their being less dispersed across the full range of specialties than are women. Women on average find that 63 percent of those graduating with them in the same major are women (P*WW), while men find that 52 percent of those they share a major with are men. These figures reflect the level of segregation as well as the fact that women represent a majority of all bachelor's degree recipients.

Does the sex segregation of majors vary across the type of college or university? The results presented in Table 15 indicate that segregation within schools is ubiquitous. It is found at roughly similar levels in all of the 8 school types in the Carnegie Foundation classification system. To the extent that segregation does vary, it is a bit less evident at Liberal Arts institutions, and a bit higher at Comprehensive Level II institutions. Selective institutions exhibited a bit more segregation than less exclusive institutions: Within each of the four major classifications, those schools classified as level I have lower levels of sex segregation than those classified as level II.

Table 16 groups schools by another set of criteria-- Historically Black versus other, and Oberlin 50 versus other. Sex segregation at Historically Black Institutions is somewhat lower than in other schools -- reflecting in part black women's choices of majors that are closer to those of white men (see discussion below). Segregation among the Oberlin 50 schools was slightly lower than in other colleges. Table 17 presents data on schools by region. Sex segregation is a bit lower in the Pacific region, but the consistency throughout the rest of the country is quite striking.

Sex Segregation: Explaining Variation Between Schools

Tables 15 to 17 are useful in providing a broad overview of the patterns of sex segregation in colleges and universities. However, they do not enable us to say a great deal about the factors that enhance or inhibit segregation at individual institutions. In this section I discuss several sets of factors that are likely to influence the level of segregation at individual institutions. I then estimate a series of regression models that estimate the effect of these factors on the segregation observed in individual colleges and universities.

A. Structure of Majors

The structure of offerings is likely to influence the level of segregation observed. For example, schools with more majors are likely to exhibit higher levels of segregation. I expect this result because more choices leaves more room for differentiation, which will probably result in higher levels of segregation between men and women. Those schools with relatively few offerings will likely display substantial overlap between men and women, simply because the choices limit the degree of segregation that is possible. The limiting case, of course, is single purpose schools, which I have excluded from this analysis.

The particular fields of study that are prominently featured at a school will also influence the degree of segregation. If an institution prominently features business as a field of study, it is likely to be relatively low in its level of sex segregation, since business nationally is a sex-neutral field of study. Education and engineering, in constrast, are more imbalanced in their sex ratios, and consequently schools where these fields are prominent will likely have higher levels of segregation.

B. Institutional Ranking

A second set of factors that might influence the level of segregation is associated with the standing of the institution. Here I have no clear prediction: historically there has been more resistance to women -- as students and faculty -- at elite institutions, but there is no special reason to believe majors are currently more differentiated within these schools. One might expect more segregation at schools with lower standing because they often have emphasize vocationally-oriented programs with links to highly sex-typed occupations. Thus, a school with a nursing program, an early-childhood education program and a variety of applied science and technology programs will likely have more segregation by sex than will a liberal arts institution. Despite the lack of clear predictions, the substantive interest in the relationship between college standing and level of segregation warrants an investigation of these factors.

As in the analysis previously discussed, I have drawn on a range of readily available measures of college standing: average SAT of the student body, acceptance rate, graduation rate, percent of faculty with PhD's, student/faculty ratio, and in-state tuition fees. For this analysis, I collapsed the Carnegie classification into five groups: liberal arts, research, doctoral granting, comprehensive and other. I also included a dummy variable to represent schools in the Oberlin 50 group. Since these factors are inter-related, I conducted a series of regression analyses with these variables included separately in order to determine which should be included in the results to be presented here. I added a dummy variable for those schools which were missing SAT data in order to include these schools in the analysis.

C. Other Institutional Factors

School size may influence segregation in many ways. Impersonal settings may minimize whatever social stigma might be associated with majoring in a sex-atypical field. If that is the case, then larger schools might display somewhat less segregation than smaller institutions.

A school's sex composition may also influence its level of segregation. Here I am interested in testing an idea that presents itself in many guises in a wide range of settings. The idea is that the presence of more women provokes a defensive reaction on the part of men, and prompts the construction of new barriers against women (Jacobs, 1992; Pfeffer and Blake-Davis, 1987). This idea is evident in the notion of the "glass ceiling" in American corporations, as well as in some prominent studies of women's entry into male-dominated occupations (Reskin and Roos, 1990). We also see it in a number of histories of women's entrance into higher education (eg., Solomon, 1985). The empirical question of interest here is whether those schools with a larger share of women graduates display a higher level of segregation.

I also explored the association between part-time enrollment and the segregation of majors. I expected that the greater the share of part-time students, the lower the social class composition of the student body and the greater the vocational orientation among students. This might lead to higher levels of segregation among the student body. Many vocationally oriented programs are highly sex-typed, reflecting the sex typing of occupations in the labor market. (Business is one of the leading exceptions at present, a field of study that is highly vocational in orientation but where both male and female students are well represented.)

The school's racial composition may also influence its level of sex segregation. As we will see, black women are less segregated from white men than are white women, and this may contribute to lower levels of segregation. A related possibility is that diversity might lead to tolerance, which might promote opportunities for both women and minorities. Finally, there is the case of historically black institutions. We expect these institutions to be less segregated by sex than are other schools, once all of the other factors discussed above are controlled, because of the choices of majors pursued by African-American women discussed below.

Finally, an institution's control structure -- public versus private -- may influence the level of segregation observed. Public institutions were among the earliest to admit women, and may be constrained from enacting restrictive practices that might be possible in private institutions. This may lead to less sex segregation by major.

Sex Segregation by Major: Regression Results

Table 18 presents a set of regression analyses designed to estimate the effect of these factors on the level of segregation among colleges and universities in the United States. There is some support for the notion that the standing of an institution affects the level of segregation. In general, the more elite schools have somewhat lower levels of segregation than less elite schools. These findings generally do not support the idea that higher status institutions continue to resist women's entrance into specific majors. Although it is well established that elite schools have been less likely to hire women faculty, they do not have higher levels of sex segregation in the fields in which they accord their degrees.

Schools with higher average SATs had somewhat lower levels of segregation. Schools with higher acceptance rates also had somewhat lower levels of segregation. A contradictory piece of evidence is the finding that schools with higher student-faculty ratios had lower levels of segregation. However, several of these school-standing effects become attenuated when other measures are included in the analysis. The average SAT score of students at an institution does continue to have a small negative effect on sex segregation, and the presumably low-SAT schools that did not report their scores were significantly higher in their level of segregation. A number of familiar measures not shown in Table 18 are not associated with the level of segregation within schools. The percent of the faculty with PhDs, the graduation rate and the institution's in-state fee level all failed to predict the level of segregation in analyses with and without other controls included, and were consequently dropped from the analyses.

I examined the effect of the type of institution in the Carnegie classification system, and a dummy variable to include whether it is one of the Oberlin-50 liberal arts schools. Oberlin-50 schools did exhibit lower levels of segregation, while a liberal-arts classification per se did not have an effect on segregation levels. Schools classified as "Carnegie Other" exhibited higher levels of segregation. This is true despite the fact that many of these are specialized institutions with a limited set of programs. However, even after the number of programs is controlled, these schools continue to exhibit somewhat higher levels of segregation. This finding supports the hypothesis that vocationally oriented programs tend to increase the level of segregation by sex. Taken as a whole, the evidence indicates that segregation levels do vary with SAT scores and other less direct measures of university standing.

Let us consider next the results pertaining to the structure of majors. As expected, both the number of fields and their specific composition affect the level of segregation. Schools

with more majors have more segregation, all other factors being equal. The larger the share of graduates in engineering and education, the higher the segregation. Conversely, the higher the fraction of degree recipients in business, the lower the segregation. (The coefficient on the engineering measure is no longer significant after all other variables are included, but the notion that the composition of fields matters is nonetheless supported throughout the analysis.)

Other institutional factors are also associated with segregation levels. Public institutions are several points less segregated than are private ones. This is consistent with our expectations. Schools with large fractions of students enrolled part-time are more segregated than others. Again, this is consistent with our discussion of vocational orientation and levels of sex segregation. The size of the student body, although it is associated with lower segregation in a number of analyses, is not significant in the final model.

I investigate the question of the relationship between the sex composition of the student body and the level of segregation in detail. When no controls are included in the analysis, there is a substantial relationship between a greater female share and higher levels of segregation. Increasing the percent female by 40 percent -- for example, from 30 to 70 percent female -- is associated with a 0.6 point increase in segregation. Once the structure of the fields of study is controlled, this drops to about 0.4 points. The percent enrolled part-time drops it to about 0.3 points. Thus, the sex composition of a college is positively related to its level of sex segregation by major, but the magnitude of this effect is small. This finding suggests that there may be a reaction in schools to a large share of women that produces higher levels of segregation by field of study. It should be noted, however, that in recent decades as the number of women in colleges increased the level of sex segregation decreased. If the cross-sectional relationship between sex composition and sex segregation holds up, then further exploration of the mechanisms responsible for this effect would be in order.

Historically Black institutions had lower levels of sex segregation. The race composition of schools did not affect the level of sex segregation once the dummy variable for Historically Black Institutions was included in the analysis.

In summary, while sex segregation is found throughout higher education, the levels found in particular schools vary. Segregation by sex across majors tends to be lower at elite colleges and universities, public institutions, and Historically Black institutions. The structure of fields matters: the more programs, the more segregation. In addition, the particular mix of offerings is important, as those schools with large gender neutral programs tend to have lower levels of segregation. Women are also more segregated from men in schools with more women and more part-time students.

Race Segregation by Major

1. African Americans

Few studies have compared the educational specialties of African-American and white college students. Trent (1991) examined changes in degrees received between 1975-76 and 1980-81. His data show that, among bachelor's degree recipients, African-American men were over-represented in education and the social sciences, and under-represented in the natural sciences and engineering, compared with white men. Other studies have also noted the relative scarcity of African Americans pursuing science and engineering (National Science Foundation, 1992). Trent found that African-American women were over-represented in business compared with white women. He notes that gender segregation dominates racial segregation in field of study: differences across race within same-sex groups were smaller than differences across sex within racial groups.

2. Levels of Segregation: African-American Women vs. White Women

A summary of national totals of degree recipients by field of study by race, ethnicity and sex across 23 specialties is presented as Table 19. (There are slightly fewer categories available for degree data by race than by sex in the HEGIS/ CASPAR data.) We will discuss African Americans first, and return to this table when we begin our discussion of Asian Americans and Hispanics. Differentiation by field of study from white women is quite modest in scope for African-American women, and is even lower for African-American men (compared with their white-male counterparts). ⁽¹⁰⁾ Fewer than one in six (14.4 percent) African-American women would have had to change fields in order to have been distributed identically to their white female counterparts (segregation indices are presented in Table 20).

Compared to racial segregation across neighborhoods, or even racial segregation across occupations, the level of segregation by field of study among undergraduate women must be judged as rather limited. Just over one quarter of African-American women would have to change occupations to be distributed in the same manner as white women (Reskin, 1994). The degree of occupational segregation between African-American women and white women has declined substantially over the course of the century (Wilkie, 1985).

The segregation between African-American women and white women is rather small even compared with other patterns of segregation among college students. As noted above, 30 percent of women would have had to change fields in order to be equally distributed with men. The data support the idea that race segregation within same-sex groups is lower than gender segregation. All of the within-sex levels of segregation are lower than all of the between-sex pairs. ⁽¹¹⁾

One intriguing aspect of the choices of majors pursued by African-American women is that their fields of study more closely match white men than do those of white women. African-American women are significantly less segregated from white men (22.5) than were white women (29.8). Which fields differentiated African-American women and white women? African-American women were over-represented in business (by 5 percentage points),

computer science and mathematics (2 percentage points), social service professions (nearly 4 percentage points), and the social sciences (nearly 2 percentage points), while white women were over-represented in the humanities (2 percentage points) and education (6 percentage points). These choices place African-American women closer to the selections of white men than is the case for white women.

It may be that African-American women are more focused on obtaining practical training that will lead to a financially secure career than are white women. The labor force participation rates of African-American women have long surpassed those of white women. Their choices of majors may in part reflect their assumption that they will be responsible for providing financial support for themselves and their families.

The under-representation of African-American women in the field of education is quite striking, because education has long been a leading field of study for African-American women. At the graduate and doctoral levels, education is by far the leading field of study for African- American women. Perhaps the drop in popularity of education reflects a combination of declining prospects for teachers and increasing opportunities in other fields.

Segregation Within Colleges: African-American Women versus White Women

The figures just discussed refer to the national totals. How does this compare to the picture observed within individual colleges and universities? A remarkable and unexpected finding is that within colleges and universities, black and white students are remarkably segregated from one another by major. Overall, more than 44.3 percent of black women would have had to change majors to be distributed in the same manner as their white classmates, despite the fact that nationally black and white women are quite evenly distributed across fields of study (see Table 20). As we will see, similar results hold for the segregation of black men versus white men by majors within schools (D=45.5).

This result is due in part to the small numbers of minority women and men obtaining degrees in many schools. In other words, the high level of segregation by race between schools discussed above poses challenges for measuring and interpreting segregation by field of study within schools. Segregation across fields by race was calculated across 23 fields of study. If there are fewer than 20 minority students obtaining degrees, then it will simply be impossible for them to be distributed across all fields of study. It is necessary to take extremely small numbers into account when calculating indices of segregation across as many as 23 units of analysis. The use of Theil's H will enable us to determine whether these high levels of segregation are simply due to random of small groups across a large set of majors.

As we saw above, the small numbers problem is quite pervasive. Of the 1554 schools, 1196 had some black female degree recipients and 1178 had some black male degree recipients. Of these, only 458 had 20 or more black women obtain degrees in 1993, and only 361 had 20 or more black men obtain degrees. In other words, there were 738 colleges with fewer than 20 black women obtaining degrees in 1993, and 817 colleges with fewer than 20 black men obtaining degrees. I did not want to exclude all of these schools from the analysis, but I did want to adjust for the fact that extremely small numbers tend to inflate the measures of segregation. My solution is to include all schools in the analysis, and to weight the descriptive results to reflect the size of the black student body. The results thus reflect the situation as measured across all schools with available data, but are weighted reflect the experiences of the average African-American college student. The tabular data in Table 20 and subsequent tables is consequently weighted by the size of the minority-group graduating class at each school, unless otherwise indicated.

The effect of small numbers becomes clear when schools are divided into those with fewer than 20 African-American women graduates and those with 20 or more. The index of dissimilarity is substantially higher for schools below this cutoff (52.4) compared to those with larger numbers of African-American women graduates (31.1). The Theil index (H) also is higher for schools with few African-American graduates. This indicates that the differences between these schools is not simply a matter of sampling variability. This comparison demonstrates the value of keeping all the schools in the analysis and using a variety of measures of segregation.

When I weighted the sample for the number of African Americans in each institution's graduating class, the index of segregation declined from 44.3 to 37.1. I then realized that I needed to remove Historically Black institutions from the analysis, because their situation was unique, as I explain below. For all schools other than Historically Black institutions, weighting the analysis reduces the level of segregation within schools between African-American women and whites, but it remains far above that observed with national data. The index of segregation calculated between African-American women and white women, weighted by the size of the black student body, is down to 28.0, but the level of within school segregation is still far higher than that observed in the national total. The H statistic also is far higher within schools than is the case nationally (H=14.2, versus 1.7 for the national total.) The level of within-school race segregation is similar to that found for within school sex segregation, despite the low national levels of race segregation by major.

African Americans are not simply segregated from whites but they are typically concentrated in a limited set of majors. The relative index of concentration is 18.3 for women, which indicates that African Americans are concentrated in a smaller range of majors than are whites. This result in part reflects the small numbers problem noted above. When the weighted average results are presented, the crowding of African-American women is reduced by about half (RC= 8.6). In terms of intergroup contact, in non-Historically Black institutions, the average African-American woman finds that 19.9 percent of the women who shares a major with her are other African-American women.

I separated out Historically Black institutions because in these schools it is the whites who are the minority, and they are the ones who are segregated into a limited set of fields of study. When all schools were included, weighting resulted in stressing the significance of these schools, and obscured the different patterns occurring at these institutions. At Historically Black institutions, I weighted the data for the number of whites, because this was the group whose small numbers posed a difficulty for the analysis. African Americans and whites are highly segregation by major within Historically Black institutions (D=45.9). However, in Historically Black institutions it is whites who are crowded into a limited set of fields of study (CR=-17.4). In essence, Historically Black institutions represent the mirror image of the experience depicted in majority-white institutions.

Despite their similar distribution in national data, within colleges African-American and white women find themselves in very different fields of study. This is evident even when I adjust for the small numbers of African-American degree recipients.

I can only speculate on the reasons for this disturbing pattern. Four possibilities are:

1) pre-college selection processes; 2) network influences on the choice of majors; 3) the influence of African-American faculty; and 4) reaction to majority group behavior. One possibility is that African-American students choose to enter certain colleges because of the reputation of specialized programs. In essence, the decision to choose the college and the major are made together. This could account for African-American students concentrating in certain majors in one college and other majors in another. However, one would have to explain why white students were not choosing the schools for the same reasons. Another possibility is that African-American students follow the choices that their classmates have made in terms of choosing courses and majors. Students surely learn about majors from their friends and classmates. The social isolation of African-American students on college and university campuses probably produces different social networks and different information about choices of majors (Allen, Epps and Haniff, 1991). In the same way that fraternity brothers steer each other to certain courses, social ties, whether based in dormitories, lunch table conversations or other social settings, could be pulling African-American students in the direction of certain majors. This again could account for the fact that African-American students are concentrated in a limited number of majors, but different ones at different schools. One piece of evidence that favors this interpretation is the finding that segregation by race is particularly high in schools with small numbers of African-American women graduates. This finding holds even when calculated with the H segregation statistic that controls for sampling variability. Small numbers of same-race peers may heighten the importance of social contacts in choosing majors.

A third possibility is that African-American students gravitate toward majors in which African-American faculty are represented. Because African-American faculty constitute such a small minority of professors, they are located in different fields in different schools. It is possible that the distribution of African-American students is influenced by these students' pursuit of same-race role models. A final possibility is that African-American students congregate in a limited set of majors because they find certain majors unwelcoming. It may be that being the only minority student is an uncomfortable situation to find oneself in. Consequently, African-American students may gravitate toward classes and majors where they find more of their same-race friends and acquaintances. This interpretation is also consistent with higher segregation in schools with few minority students.

I don't believe these results could be due to differential failure rates in highly competitive fields. If that were the driving force behind segregation between African-American and white students, then African Americans would be found in the same majors no matter which schools they attended. Instead, the pattern that we observe is high levels of segregation by race within schools, but the clustering occurs around different majors on different campuses.

Whatever its cause, I find this pattern disturbing because it raises questions about students' ability to make the best match between their talents and the fields they may study and the careers that they may be able to enter. Having choices effectively limited is not a desirable state of affairs, even if active discrimination is not the driving force that causes this pattern. The academic isolation of minority students is a topic that calls for much further research.

African-American Men versus White Men

Which fields account for the differentiation of African-American men from white men?

This question can be addressed with the data presented in Table 19. African-American men were under-represented in engineering (3 percentage points) and other sciences except computer science, where African-American men surpass whites by one percentage point. African-American men were over-represented in the social sciences and social-service professions (which includes law enforcement), while white men were over-represented in agriculture and the humanities. In general, I am struck by the large number of fields in which African-American men and white men were represented equally: business, communications, education, and psychology, among others.

The level of differentiation by majors between African-American men and white men is modest: nationally, only 9.2 percent of African-American men would have to change majors to be distributed in the same way as white men (see Table 21). This level is lower than that found between African-American women and white women, and is far lower than the occupational segregation found between African-American and white men. Recent studies of the

U. S. economy indicate that nearly one third of African-American men would have to change occupations and over half would have to change job titles to be distributed in the same manner as white men (Reskin and Cassirer, 1997; Tomaskovic-Devey, 1993).

The balance of the results presented in Table 21 for African-American men parallel those presented in Table 20 for African-American women. The extent of segregation is markedly higher in those schools where there are fewer than 20 African-American men graduating. The weighted results reduce the measured level of segregation, as was the case for African-American women. The level of within-school segregation by majors far exceeds the national totals for the two principal measures of segregation (D and H). Among men, segregation is quite high by major in Historically Black institutions, with whites concentrated into a limited set of majors.

Asian Americans

Asian Americans have the most distinctive pattern of major choice of the minority groups discussed in this report, although the Asian American versus white differences still remain smaller than do differences in the choice of major by sex (see Table 19). Asian-American women are more concentrated in science and engineering than are white women. They lead white women in engineering (5 percentage points), computer science (3 percentage points), biological sciences (6 percentage points), and business (7 percentage points), while they trail white women in education (10 percentage points), the humanities, and arts and music. As a result, Asian-American women fall much closer to white men's choices of majors than do white women (D=18.9, versus 29.8 for white women versus white men). Nearly one quarter (23.4 percent) of Asian-American women would have to change majors to be distributed in the same manner as white women.

With this national picture as the backdrop, how much differentiation is there between white women and Asian-American women by major within schools? The results of this analysis are presented in Table 22. We see the same pattern of high levels of within-school segregation by major for Asian-American women versus white women that we saw in the case of African Americans. The within-school segregation by majors far exceeds that observed in the national data. The (weighted) level of segregation within schools is higher (33.0) than that observed nationally (23.4). The H statistic that adjusts for random variability in small samples also shows much more segregation within schools than is the case nationally.

Asian-American men are the most concentrated of any group in male-dominated fields of study. One of five (20.7 percent) obtains his degree in engineering, and Asian-American men exceed whites in every natural scientific discipline listed in Table 19. They trail white men in majoring in business and in education. Nationally, just under one quarter (23.6 percent) would have to change majors to be distributed in the same manner as whites (see Table 22).

The within-college story for Asian-American men parallels that for Asian-American women. There is substantially more segregation by major within colleges and universities than is observed nationally. This result holds even when the data are weighted to avoid the bias due to small sample sizes, and holds for the H statistic that adjusts for sampling variability.

Hispanics

Nationally, Hispanics major in the same subjects as whites. Only 13.5 percent of Hispanic women and 6.8 percent of Hispanic men would have to change majors to be distributed in the same fields as their same-sex white counterparts (see Table 19 for specific fields and Table 23 for summary statistics). This represents a level of segregation similar to that found for the case of African Americans. For Hispanics, as was the case for African Americans and Asian Americans, segregation between Hispanics and non-hispanic whites within schools is very high. For Hispanics, the levels of within-school segregation are approximately as high as is the case for African Americans and Asian Americans.

Summary

Minority groups vary in the choices of majors, yet several important commonalities are evident for all minority groups. First, gender is more important than race or ethnicity in the choice of major. Women from different race and ethnic groups are generally more similar in their choices of major than are women with men from their same ethnic group. Second, within schools, minority groups find themselves much more segregated by major from whites than they are nationally. The excess of within-school over national differentiation is quite dramatic for all groups, and is robust across different measures and different ways of manipulating the data. It is an intriguing and disturbing result that calls for further investigation.

Regression Results

What factors are associated with high levels of race and ethnic segregation by major within schools? Are these the same factors that are responsible for high levels of sex segregation within schools? Descriptive tables for each minority group by sex are presented as Appendix Tables 2-19. I focus my discussion in the text on the regression results presented in Table 24.

I regressed a number of factors on the level of race and ethnic segregation within schools, following the logic developed in the analysis of sex segregation between schools presented in Table 18. The first set of measures are indicators of school ranking or quality; the second taps the offerings of schools. The third set measures a variety of an institution's attributes, such as its public status, and measures of the social composition of the student body.

In summarizing the results presented in Table 24, I would like to stress the factors that tend to have common effects across most if not all of the groups. There tends to be less segregation in high status institutions. For example, the effect of average SAT score is negative in all six equations, and is statistically significant in five of the six. On the other hand, the effects of acceptance rate and student faculty ratio are not consistent across the groups.

Institution type effects are often significant but not entirely uniform. Research and doctoral granting institutions typically exhibit lower levels of race and ethnic segregation by major, compared to comprehensive institutions, as do schools classified as "other." Liberal arts institutions usually have higher levels of race and ethnic segregation, but, within the liberal arts colleges, the Oberlin 50 usually have lower levels of segregation.

The specifics of program offerings effect the levels of segregation by major for all groups. The larger the share of students obtaining their degrees in education, the higher the segregation by major by race and ethnicity. The larger share in business, the lower the segregation observed. Engineering tends to follow the same pattern as education, but the effects are less consistent. The more majors offered, the more segregation results.

Public institutions on average have lower race and ethnic segregation than do private schools. The more part-time students, the lower the level of segregation. This is one clear case in which race and ethnic segregation by major tend to have a very different pattern than does gender segregation.

Segregation generally declines as the percent of minority students increases. There are two apparent exceptions to this generalization, but upon closer inspection they fall into line. The exceptions are for African-American and Hispanic men. For African-American men, the positive relationship is due to high levels of segregation in Historically Black institutions, resulting from the skewed distribution of a small number of white men. When this group of schools is excluded or controlled, then a negative relationship between segregation levels and minority presence appears. The same logic holds for the Hispanic case. The levels of major segregation by ethnicity are very high in Puerto Rico, stemming from the distribution of a handful of non-Hispanic whites. When a dummy variable for Puerto Rico is added to the equation, a negative relationship between minority composition and segregation levels becomes evident.

Regional effects are generally consistent across groups. Schools in the Midwest and South usually have higher levels of segregation than those in the Northeast, while the West is

often the lowest region of all.

Discussion

I began with the premise that gender operates quite differently than race and ethnicity in the context of higher education. Given the surge in women's enrollments, gender inequality no longer takes the form of differences in access. If women are at a disadvantage in college, it must be in terms of the majors they select or the schools they attend. I expected that women and men were quite evenly distributed across colleges but segregated by field of study within colleges. In contrast, African Americans continue to lag in terms of access to higher education. Differences in the college experience would compound these initial differences. Based on my earlier exploration of the distribution of majors of African Americans, I expected that minorities would be segregated between schools but close to evenly distributed by majors within schools.

These initial expectations received substantial support, but they need to be modified in several key ways. First, women are not quite evenly distributed across institutions. Women remain under-represented in high status institutions. The gender segregation that remains is principally due to two factors: 1) women's under-representation in schools that feature large engineering programs, and 2) women's over-representation among part-time students. Within- school segregation by sex is substantial, but it varies across institutions. Segregation is generally lower at higher status institutions, a finding that contradicts the expectation derived from a number of studies in diverse contexts. The structure of course offerings is a prime driver of the level of segregation. Schools with more programs, and those featuring highly segregated majors are most likely to be segregated. Vocationally-oriented programs such as nursing and criminal justice that are linked to highly sex-typed occupations tend to be highly segregated by sex.

The findings with respect to race and ethnicity conformed to expectations in several ways but revealed a number of surprises as well. Substantial segregation by race and ethnicity is evident in institutions of higher education in the U. S. I expected this pattern for African Americans, given the continuing high enrollment of African-American students in Historically Black institutions. However, Asian-American and especially Hispanic students also are unevenly distributed across institutions. For these groups, an uneven geographic distribution contributes to segregation between schools. African-American and Hispanic graduates trail whites in the status of the schools from which they obtain their degrees, while Asian-American students graduate from higher ranking schools than do whites.

The distribution of minorities across majors is quite similar to that of whites nationally.

For African-American and Hispanic students, the differences in the choice of majors vis-a-vis white students are modest. Asian-American graduates are most distinctive in their choices of majors.

However, a major surprise of this study has been that minority students are highly segregated by major from their white colleagues within specific colleges and universities. This pattern was evident for African-American, Asian-American and Hispanic students, and is also found for white students studying in Historically Black institutions.

My initial expectations stressed the differences between gender on the one hand and race and ethnicity on the other. However, the determinants of segregation within schools suggest strong commonalities across these groups. In other words, many of the same factors that promote gender equity also tend to promote less segregation of minorities.

This research has identified a number of striking and intriguing findings. More research on the causes and consequences of the gender, race and ethnic segregation of American colleges and universities is in order to explain the patterns documented here.

References

Allen, Walter R., Edgar G. Epps, and Nesha Z. Haniff, eds., 1991. College in Black and White. Albany, NY: State University of New York Press.

Hearn, James C. 1990. "Pathways to Attendance at the Elite Colleges." pp. 121-146 in Paul W. Kingston and Lionel S. Lewis, eds., The High-Status Track: Studies of Elite Schools and Stratification. Albany, NY: SUNY Press.

Karen, David. 1991. "The Politics of Class, Race and Gender: Access to Higher Education in the United States, 1960-1986." American Journal of Education 208-237.

Jacobs, Jerry A. 1996. "Gender Inequality and Higher Education." Annual Review of Sociology.

22:153-85.

Jacobs, Jerry A. 1995. "Gender and Academic Specialties: Trends Among Recipients of College Degrees During the 1980s." Sociology of Education 68(2):81-98.

Jacobs, Jerry A. 1993."Theoretical and Measurement Issues in the Study of Sex Segregation in the Workplace," European Sociological Review 9(3):325-330.

Jacobs, Jerry A. 1992. "Women's Entry into Management: Trends in Earnings, Authority, and Values Among Salaried Managers," Administrative Science Quarterly 37(2): 282-301.

Jacobs, Jerry A. 1989. Revolving Doors: Sex Segregation and Women's Careers. Stanford

University Press.

Jacobs, Jerry A. and Suet T. Lim. 1992. "Trends in Occupational and Industrial Sex Segregation in 56 Countries, 1960-1980," Work and Occupations 19(4):450-486. Reprinted in Jerry A. Jacobs, ed., Gender Inequality at Work. 1995. Beverly Hills, CA: Sage Publications.

Massey, Douglas S. and Nancy A. Denton. 1988. "Dimensions of Residential Segregation." Social Forces 67(2)281-315.

National Center for Educational Statistics. 1995. Digest of Educational Statistics. Washington, D. C.: U. S. Government Printing Office.

Persell, Caroline Hodges, Sophia Catsambis, and Peter W. Cookson. 1992. "Differential Asset Conversion: Class and Gendered Pathways to Selective Colleges." Sociology of Education 65:208-225.

Pfeffer, Jeffrey and Alison Davis-Blake. 1987. "The Effects of the Proportion of Women on Salaries: The Case of College Administrators," Administrative Science Quarterly 32:1-24.

National Science Foundation. 1992. Women and Minorities in Science and Engineering: An Update. Washington, D. C.: National Science Foundation.

Reskin, Barbara F. and Patricia Roos. 1990. Job Queues, Gender Queues. Philadelphia: Temple University Press.

Reskin, Barbara F. and Naomi Cassirer. 1997. "Occupational Segregation by Gender, Race and Ethnicity." forthcoming, Sociological Focus.

Solomon, Barbara Miller. 1985. In the Company of Educated Women: A History of Women in Higher Education in America. New Haven: Yale University Press.

Theil, Henri. 1972. Statistical Decomposition Analysis. Amsterdam: North Holland.

Theil, Henri and Anthony J. Finizza. 1971. "A Note of the Measurement of Racial Integration of Schools by Means of Informational Concepts." Journal of Mathematical Sociology 1:187-194.

Tomaskovic-Devey, Donald. 1993. How Come Everybody Looks Like Me: The Causes and Consequences of Sex and Race Segregation at Work. Ithaca, New York: Industrial and Labor Relations Press.

White, Michael J. 1986. "Segregation and Diversity Measures in Population Distribution." Population Index 52(2):198-221.

Wilkie, Jane R. 1985. "The Decline of Occupational Segregation between Black and White Women. Research in Race and Ethnic Relations. 4:67-89.

Trent, William T. 1991. "Focus on Equity: Race and Gender Differences in Degree Attainment," 1975-76; 1980-81" pp. 41-60 in Walter R. Allen, Edgar G. Epps, and Nesha Z. Haniff, eds. in Black and White: African-American Students in Predominantly White and Historically Black Public Universities. Albany: State University of New York Press.

U. S. Bureau of the Census. 1996. Data obtained from the U. S. Census World Wide Web

Site. Figures are based on March 1994 Current Population Survey.

Table 1. Admission of Women to Leading Universities

and Liberal Arts Colleges, Fall 1995

Source: Monthly Forum on Women in Higher Education,

October 1995.

1. HEGIS stands for the Higher Education Graduates Information System.

2. CASPAR stands for the Computer Aided Science Policy Analysis and Research Database System. Prepared by the Quantum Research Corporation of Bethesda, MD, CASPAR makes HEGIS data available through an on-line computer server system.

3. There are several definitions of two-year schools in the HEGIS/ CASPAR data, based on "Institutional Class" or "Highest Degree Awarded" in addition to the Carnegie definition. These definitions vary substantially.

4. I want to thank my colleague Kermit Daniel of the Wharton School of Management at the University of Pennsylvania who generously provided me with these data.

5. The data include all degree recipients, and consequently the standard issue of generalizing from a sample to the population does not apply here. However, in the multi-variate analyses I treat the data as a sample with 1500 cases and use conventional measures of statistical significance as one measure by which results can be judged to be substantively important.

6. I conducted additional analyses not reported here to determine whether the differences between the results reported here and those of Hearn and Persell are due to changes over time in women's standing or to differences in methodology. I found results similar to those presented here for 1980 and 1970, which suggests that engineering and part-time enrollment have long been responsible for women's under-representation in top-tier institutions.

7. Data were missing for 18.6 percent of African Americans compared to 10.5 percent of whites. Since schools with missing data probably are below average in rankings, the race gap in school average SAT is probably even bigger. The results presented here substitute the mean SAT score for missing data. This procedure tends to minimize group differences by inflating the average for the disadvantaged group. I found that school size was the best predictor of missing SAT data: small schools were less likely to report information than big ones. Once school size is controlled, there is no relationship between race composition and the likelihood of missing data on SAT scores.

8. In the logistic regression equations for representation in schools with average SATs over 1200 and 1100, the dummy variable for Historically Black institutions could not be included, because of high collinearity with the dependent variable.

9. I explored whether regional controls affected the results reported for African Americans and Asian Americans. While region is a very powerful predictor of SAT scores and many other measures, the results reported for these groups do not change substantively with the addition of regional controls.

10. (Elsewhere I have shown that the level of segregation by field of study declined substantially during the period 1977-1991 between African-American men and white men, but changed little between African-American women and white women. The level of segregation is markedly higher among master's and doctoral degree recipients because of the greater concentration of African Americans in the field of education among master's and doctoral degree recipients. This final result suggests significant differences in the persistence rate between college and graduate school for African-Americans in particular specialties.

11. The level of sex segregation by majors for African Americans was 19.2, which was higher than that found for African-American women versus white women (14.4) and African-American men versus white men (9.2). Similarly, sex segregation among Hispanics (22.7) far exceeds that found between Hispanic women and white women (3.0) and Hispanic men and white men (1.8). Only in the case of Asian Americans does race segregation within sex approach the level of sex segregation within race. Sex segregation among Asian Americans is 24.4, compared with racial segregation (Asian Americans versus white) of 23.4 for women and 23.6 for men.