CHAPTER 5

Career and Economic Impacts of College

 

            In this chapter we summarize the accumulated evidence on the career and economic impacts of college.  Overall, the labor market impacts of postsecondary education have remained substantial during the decade of the 1990s; and this fact has not been lost on students about to graduate from high school.  For example, of the 2.7 million American high school graduates in 1996, nearly two-thirds (65%) went on to some type of postsecondary education (College continuation rates for recent high school graduates reached record high in 1996, 1997, August). Moreover, the reasons for attending college appear to be strongly linked to a perception that a college degree gives one decided economic and career advantages.  According to the 1997 survey of American college freshmen by the Higher Education Research Institute at UCLA, nearly three-fourths said that getting a better job (74.6%) and making more money (73.0%) were the most important reasons for attending college (see also, Flacks & Thomas, 1997; Is college still worth the cost?  The private investment value of higher education 1967 to 1996, 1998, March).  We examine these labor market or occupational attainment impacts in considerable detail, but we also look at the evidence pertaining to other dimensions of career such as career maturity, career progression and success, and job satisfaction.

            By definition, of course, a large part of the evidence on the career and economic impacts of college concerns long-term or extended impacts.  Consequently, we do not include a separate section on long-term effects in this chapter.  Rather, we synthesize the evidence on such effects, where appropriate, within our other five sections.


Change During College

Conclusions from How College Affects Students

            A consistent body of evidence suggests that students become significantly more mature, knowledgeable, and focused during college in thinking about planning for a career.  Whether this is an effect of college or simply a development that occurs coincidentally with college attendance is difficult to determine.  The simple fact of having to confront one’s life work may have a substantial impact on the increased maturity found in seniors’ thinking and planning for a career.

Evidence from the 1990s

            The small body of evidence from the 1990s is quite consistent with our 1991 conclusion that students become more mature, knowledgeable, and focused during college in thinking about a career.  For example, Luzzo (1990; 1993) sought to determine if undergraduate class standing was related to scores on a standardized measure of career maturity.  Career maturity was defined as the readiness of an individual to make informed, age-appropriate career decisions and cope with appropriate developmental tasks (Savickas, 1990).  Although the statistically significant relationship was not particularly strong, Luzzo found that the more advanced one’s undergraduate class standing, the higher one’s level of career maturity.  While they employ different measures than Luzzo, similar results have been reported by Bowman and Tinsley (1991) for vocational realism, by Poe (1991) for stability of vocational identity and need for occupational information, and by VanHaveren, Winterowd, and Fuqua (1999) for career decidedness.  Furthermore, there is also evidence to suggest that college seniors have a more accurate perspective about labor-market realities than do students in the first two years of college.  Two studies indicate that seniors have significantly lower and more realistic estimates of actual starting salaries for college graduates than either freshmen (Heckert & Wallis, 1998) or sophomores (Shepperd, Oullette, & Fernandez, 1996).

            Despite the consistent evidence, interpreting such differences in career maturity, career identity, and vocational/labor-market realism as an impact of college is problematic.  Simple maturation or, as we suggested in our previous synthesis, the increased pressure on seniors to reach closure on career decisions may be equally valid as competing explanations for the findings.

            In addition to focusing on the different dimensions of career maturity or career identity, other research has examined the extent to which college students are prepared to meet workplace requirements.  This work was conducted by the Collegiate Employment Research Institute at Michigan State University (Gardner, 1998).  A simulation-based assessment of a real workplace situation was developed that tapped individuals’ competencies in such areas as applied problem solving, interpersonal effectiveness, and accountability.  The instrument was normed on a group of recent college graduates evaluated as performing above average on a common job-assessment instrument.  Consistent with the evidence on career maturity, college seniors tended to have the highest overall workplace readiness scores, while freshmen had the lowest.  Gardner reports that the class-level differences in workplace readiness scores were modest, although no standard deviations were reported on which an effect size could be estimated.  Irrespective of the size of the class-level differences in workplace readiness, however, attributing them to the college experience itself is hazardous.  Since the study is cross-sectional in design, factors such as age and mortality of the least able students from upper-class samples might also explain the class-level differences.

            The Gardner (1998) study not only reports the relative standing of freshmen, sophomores, juniors, and seniors in workplace readiness, it also provides an absolute estimate.  As alluded to above, the workplace readiness assessment produced a normed score for employed college graduates who have been evaluated as performing above average on the job.  This normed score was substantially above that of the average college seniors, suggesting that, while they are more proficient than those with less exposure to college, seniors may often lack an absolute level of job skills required for above average job performance.  Corroborating evidence is suggested in an additional survey of employers who first specified the absolute skill or performance levels new graduates should be expected to meet upon entry into their jobs, and then indicated the level of preparedness of college graduates they observed.  While students generally appeared to be well prepared in their academic and content areas, they fell short in areas that were related to the context of work (e.g., interpersonal skills, setting priorities) and applying their knowledge in work environments (Gardner, 1998; see also Candy & Crebert, 1991; O'Brien, 1997; Van Horn, 1995).

Net Effects of College

Conclusions from How College Affects Students

            Attaining a bachelor’s degree has important implications for the type of job one obtains and for an individual’s lifetime earnings.  Our best estimates were that, net of an individual’s background and other confounding influences, a bachelor’s degree (compared to a high school diploma) conferred about a 34 percentile point advantage in occupational status or prestige, a 20 to 40% advantage in earnings, and a private rate of return of between 9.3 and 10.9%. The occupational status advantage that accrued to college graduates was not simply a function of the first job obtained.  Rather, the significant occupational status differences between high school and college graduates were sustained over the occupational life span, even when the status of one’s first job is taken into account.  For both occupational status and earnings, there was a credentialing effect.  One received a “bonus” for completing the bachelor’s degree above and beyond the increment in job status or earnings received for every year of postsecondary education.

            The fact that a bachelor’s degree significantly enhanced the likelihood of entering relatively high status managerial, technical, and professional occupations has implications not only for earnings, but also for occupational stability.  The very nature of jobs entered by college graduates (versus high school graduates) tended to make them less sensitive to employment fluctuations that occur with changing economic conditions.  This may at least partially explain why college graduates were substantially less likely than high school graduates to be unemployed.  Related evidence also suggested that college-educated individuals may have additional hedges against prolonged periods of unemployment in the form of increased accuracy of occupational information and efficiency in job search, increased regional mobility to take advantage of employment opportunities, and an increased network of personal contacts, some of which date back to college days.

            Despite the fact that one derived substantial occupational status and earnings advantages from a bachelor’s degree, irrespective of his or her background characteristics, the causal mechanism(s) underlying the ability of a bachelor’s degree to confer these advantages was (were) not readily apparent.  On the one hand, we found evidence to support a socialization or human capital explanation.  That is, college imparts cognitive skills, values, attitudes, and behavioral patterns that make the individual more productive in complex technical, professional, and managerial occupations, and therefore more highly paid.  On the other hand, we also found evidence supporting a screening or certification explanation.  This explanation posits that a college degree serves a screening or certification function such that those without a bachelor’s degree are effectively barred from entry into high status/high income careers.  If college-educated individuals are perceived by employers as more likely than high school graduates to possess the requisite competencies and values necessary for successful adaptation to complex technical and managerial positions, they will continue to secure higher-status and better paying jobs irrespective of whether the competencies and values were acquired in college.  Our reading of the total body of evidence was that both socialization/human capital and screening/certification may be part of any causal link between postsecondary education and both occupational status and earnings.  Neither hypothesis alone provided a completely satisfactory explanation.

            College tended to produce conflicting influences on satisfaction with one’s work.  On the one hand, college tended to have a modest positive influence on job satisfaction by placing individuals in jobs with relatively high intrinsic (autonomy, challenge, interest) and extrinsic (income) rewards.  On the other hand, college tended to develop a capacity for critical judgment and evaluation that may make college-educated individuals more sensitive to the shortcomings of their jobs.  Similarly, the college-educated were also quite likely to have higher expectations about the intrinsic rewards of their jobs than those with less education.  The latter factors can lead to dissatisfaction in situations of “overeducation,” where job demands do not require a college-level education.

Evidence from the 1990s

            Evidence from the 1990s on the net career and economic impacts of college differs somewhat in focus from the evidence we reviewed from our 1991 synthesis.  In our present synthesis we found substantially less evidence on such topics as occupational productivity, job satisfaction, and job success, but substantially more on the economic returns to different levels of postsecondary education.  In reviewing this evidence, we have had the benefit of several excellent literature reviews (e.g., Boesel & Fredland, 1999; Grubb, 1998, August; Paulsen, 1998), which has made the locating of studies and the development of this chapter substantially easier.  Still, the body of evidence is extensive.  Thus, in an attempt to provide some organizing structure we present the evidence in the following categories: occupational status, workforce participation, job satisfaction/performance, earnings, credentialing effects, private rate of return, and causal mechanisms.

Occupational Status

            Occupational status can be generally regarded as a hierarchy of occupations that reflects their prestige or desirability.  Not surprisingly, perceived occupational prestige or desirability in the United States has an overwhelming socioeconomic basis consisting largely of education and income (Stevens & Featherman, 1981).  Typically, an occupation’s status, or Socio-Economic Index (SEI) score, depends on the percentage of individuals working in that occupation who have completed a certain level of formal education or higher and the percentage with incomes at a certain level or higher.  Since an occupation’s SEI score is a function of its educational requirements, it would seem, on first consideration, that any association between formal education and occupational status is largely tautological.  This probably overstates the case, however.  The link between formal education and occupational status reflects a real social phenomenon that has implications for the cognitive complexity and desirability of the work, the social position of those who engage in the work, and their children’s life chances (Jencks et al., 1979; Pascarella & Terenzini, 1991).

            We uncovered only a small body of studies published in the 1990s that focused on the impact of postsecondary educational attainment on occupational status.  The results of this research are quite consistent with the evidence from our previous synthesis, but the data sets analyzed are somewhat dated.  Two studies (Knox, Lindsey, & Kolb, 1993[1]; Lin & Vogt, 1996) analyzed the 1986 follow-up of the National Study of the High School Class of 1972 (NLS-72), and one study (Kerckhoff & Bell, 1998) used the 1986 follow-up of the 1980 High School and Beyond (HS&B) data.  An additional study by Lavin and Hyllegard (1991) analyzed a single-institution sample of 1970-72 graduates followed up in 1984.  What the evidence from these studies suggests is that: 1) a bachelor’s degree provides occupational status advantage over a high school degree of about .95 of a standard deviation (33 percentile points); 2) an associate’s degree confers an estimated occupational status advantage over a high school degree of between .24 and .44 of a standard deviation (9 to 17 percentile points); and 3) other amounts of postsecondary education or sub-baccalaureate credentials such as a vocational degree or a license/certificate provide an estimated occupational status advantage over a high school diploma of between .12 and .22 of a standard deviation (5 to 9 percentile points).  All of these occupational status advantages over those conferred by a high school degree are statistically significant and persist even in the presence of statistical controls in different studies for such influences as family socioeconomic status, sex, race, high school achievement, standardized test scores, educational aspirations, and occupational aspirations.[2]  Thus, while a bachelor’s degree appears to confer the largest incremental advantage in occupational status compared to a high school diploma, the evidence from the 1990s suggests that various forms of sub-baccalaureate post-secondary education also provide modest but statistically significant occupational status advantages.

Workforce Participation

            Consistent with the conclusions from our previous synthesis, evidence from the 1990s clearly suggests that the bachelor’s degree not only increases the likelihood of entering relatively high status technical and professional occupations, it also increases the likelihood of holding a job that provides relative occupational stability.  Compared to high school graduates, those with a bachelor’s degree are substantially more likely to participate in the labor force and are substantially less likely to be unemployed (Blau, Ferber, & Winkler, 1998; Grubb, 1998, August; Office of Education Research and Improvement, 1996; Paulsen, 1998; U.S. Department of Education, 1992; Veum & Weiss, 1993).  For example, data from the U.S. Bureau of the Census indicated that in 1996 about 84.7% of all bachelor’s degree holders age 25-64 were employed, or participated in the workforce, while the corresponding rate for those with a high school diploma was 74.6%.  Conversely, the 1996 unemployment rate for bachelor’s degree holders was 2.4%, less than half that for those with a high school degree, 4.7% (Employment and unemployment rates by educational attainment 1970 to 1996, 1997, August).

            Although they are less dramatic in magnitude, similar differences exist between high school graduates and those with sub-baccalaureate degrees or credits in postsecondary education (Blau et al., 1998; Grubb, 1998, August; Veum & Weiss, 1993).  Using the same 1996 Census data, the workforce participation rates were: 82.3% for those age 25 to 64 with an Academic Associate’s degree, 83.9% for those with a Vocational Associate’s degree, 79.4% for those with some postsecondary education credits but no degree, and 74.6% for those with a high school diploma.  The unemployment rates were: Academic Associate’s degree = 3.2%, Vocational Associate’s degree = 3.3%, some postsecondary education credits but no degree = 4.0%, and high school diploma = 4.7% (Employment and unemployment rates by educational attainment 1970 to 1996, 1997, August).

            Of course, such findings do not necessarily reflect the net impact of a postsecondary degree completed or postsecondary credits taken.  Rather, they may simply represent employability-related differences in the intellectual or personal characteristics of individuals with varying levels of formal education.  We uncovered two studies that addressed this issue with nationally representative data.  Lewis, Hearn, and Zilbert (1993) analyzed the 1986 follow up of the High School and Beyond data base to determine if postsecondary vocational training influenced participation in the labor force and number of months employed.  Net of controls for such factors as race, gender, academic ability test scores, secondary school grades, socioeconomic background, and high school program, neither participation in vocational training nor completing a vocational training program significantly influenced either measure of workforce participation.  Consistent findings are reported by Surette (1997) in analyses of 12 years of data (1979-1990) for men from the National Longitudinal Study of Youth.  Net of controls for ability test scores, labor market experience, and age, the completion of sub-baccalaureate vocational training had only small and nonsignificant effects on both the probability of being employed and annual hours worked.  However, both the number of community-college credits completed, and the number of four-year college credits completed had a significant positive effect on the probability of employment.  The number of four-year college credits completed and the completion of a bachelor’s degree had significant positive effects on annual hours worked.  Thus, there is a modicum of support for the hypothesis that the relationship between amount of formal postsecondary education and workforce participation is causal and not simply attributable to the characteristics of individuals who acquire different amounts of postsecondary credits or degrees.

Job Satisfaction/Performance

            Although it is possible that we may have missed some studies, our literature search for the present synthesis uncovered relatively little evidence on job satisfaction and job performance.  What evidence we did uncover, however, is consistent with our previous synthesis in suggesting that postsecondary education tended to produce conflicting, or at least complex influences on satisfaction with one’s work.  It is clear that having a college degree increases the likelihood that one will be engaged in work that not only provides higher levels of extrinsic rewards (e.g., prestige and income), but which also offers greater intrinsic rewards (e.g., complexity, autonomy, managerial authority, ideational content, nonroutine tasks, and sense of control over one’s work) (Grubb, 1998, August; Hyllegard & Lavin, 1992; Kohn et al., 1990; Ross & Reskin, 1992).  Indeed, net of academic ability, socioeconomic background, race, and gender, increasing one’s level of postsecondary education appears to increase the importance of such intrinsic work values (Knox et al., 1993).  Having a college education tends to have a positive indirect effect on job satisfaction through its impact on such factors as job prestige and earnings, job autonomy, and nonroutine work.  However, net of those factors, the direct effect of having a college degree on job satisfaction tends to be negative, possibly because education functions to raise workers’ expectations (Ross & Reskin, 1992).  There is also evidence indicating that when college graduates hold jobs that do not typically require a college degree, such “over-education” can have a negative effect on job satisfaction (Jenkins, 1992).  Part of this negative effect may be because college graduates have higher expectations of the intrinsic characteristics or returns of work than those of their actual job (Jenkins, 1992), while another part may be attributable to the negative influence of over-education on the extrinsic rewards of work, such as earnings (Verdugo & Verdugo, 1989).

            We also uncovered little research on the influence of postsecondary education on job performance.  Hill (1989) surveyed nearly 190 employers in Pennsylvania to determine the effects of postsecondary education on the performance of over 500 employees in six technical occupations: computer programmers, EDP equipment operators, electrical/electronics engineering technicians, mechanical engineering technicians, drafters, and surveying technicians.  In the employee sample, 32% had a high school degree, 17% had a bachelor’s degree, and 51% had some postsecondary education.  With statistical controls for the number of employees in the company and the type of industry, workers with some postsecondary education or a bachelor's degree tended to display statistically significant performance advantages—performing better when starting work and requiring a shorter training period.  They were also more likely to be promoted.  While such evidence suggests that postsecondary education improves job performance, it should be cautioned that the job classifications in the Hill study often do not typically require a bachelor’s or even an associate’s degree.  Whether the same results would hold in higher level managerial or professional positions is not clear.  We uncovered little consistent evidence in our 1991 synthesis to suggest job productivity differences when college-educated and noncollege-educated individuals hold the same job, although the former may have greater career mobility.  Furthermore, it is difficult to attribute the findings of Hill’s study to the influence of postsecondary education.  Since no controls were made for employee background characteristics, the findings could just as easily be attributed to differential recruitment.  As compared to their counterparts with high school degrees, those with exposure to postsecondary education may simply possess more of the personal characteristics that contribute to effective job performance to begin with.

Earnings

            Earnings in the United States are strongly related to level of formal schooling or educational attainment.  In the decade of the 1990s, a series of important studies has attempted to determine if that relationship is causal or simply the result of individuals with greater personal earning capabilities (e.g., intellectual ability) attaining higher levels of formal schooling.  This research has tended to use either samples of identical twins (Ashenfelter & Krueger, 1994; Ashenfelter & Rouse, 1998a; Rouse, 1999) or naturally occurring randomized experiments such as age at the beginning of compulsory school attendance or the military draft lottery (Angrist & Krueger, 1991, 1993).  As suggested by Ashenfelter and Rouse (1998b), the synthesis of these studies is remarkably consistent; it indicates that the earnings return to formal schooling is quite likely causal and is not attributable to the omitted correlations between schooling and personal attributes such as intellectual ability.  Depending on the analytical approach taken, the typical economic return to each year of formal schooling completed is an advantage (or premium) of somewhere between 5% (Ingram & Neuman, 1999) and 16% (Ashenfelter & Krueger, 1994), with an average of about 9-10% (Ashenfelter & Rouse, 1998a; Rouse, 1999).  Of course, this estimate is an average based on each additional year of schooling completed.  Different levels of schooling and different credentials and degrees may vary in the net earnings premium they return to the individual.  A substantial body of research in the 1990s has attempted to estimate the earnings premium to different postsecondary degrees and credentials, as well as to years of postsecondary education when a degree or credential is not completed.

            If one considers the premium to a bachelor’s degree simply as the average earnings of individuals with a bachelor’s degree relative to the average earnings of those individuals with a high school degree, expressed as a percentage, then it is reasonably clear that the premium to a bachelor’s degree in the United States has increased during the last part of the twentieth century (Boesel & Fredland, 1999; Bound & Johnson, 1992; Freeman, 1994; Grogger & Eide, 1995; Katz & Murphy, 1992; Levy & Murnane, 1992; Murphy & Welch, 1992a; Pencavel, 1991).  This increase is clearly illustrated in Current Population Survey data from the Census Bureau for the average annual earnings of men and women 25 and older (Is college still worth the cost?  The private investment value of higher education 1967 to 1996, 1998, March).  For the five-year period of 1967-71, male and female bachelor’s degree holders had an average annual earnings advantage (unadjusted for inflation) of 48.5% over their counterparts with a high school diploma.  (In other words, for this five-year period the average annual earnings of those with a bachelor’s degree was 1.485 times as large as those with a high school diploma.)  In contrast, for the five-year period 1992-96, men and women with a bachelor’s degree had an average annual earnings advantage of 79.8% over men and women with a high school degree.  The only aberration in this steady increase over time in the college premium has been a downtrend in the 1970s, which paralleled the arrival of the “baby boomer” cohorts into the U.S. labor market (Berger, 1989; Boesel & Fredland, 1999; Murphy & Welch, 1992a, 1993).  Indeed, for the five-year period from 1974-79 college graduates as a group were only earning 43% more than those with a high school diploma (Is college still worth the cost?  The private investment value of higher education 1967 to 1996, 1998, March).[3], [4]

            Of course it is doubtful that the total earnings premium associated with a bachelor’s degree (versus a high school degree) is entirely attributable to college attendance.  Compared to high school graduates, individuals who attend and graduate from college may simply possess more of the cognitive skills and personal attributes that lead to success and high earnings in complex managerial and technical jobs to begin with.  A body of research in the 1990s has attempted to estimate the net earnings premium of a bachelor’s degree (versus a high school degree) by introducing various statistical controls for differences among individuals that might confound the relationship between level of formal education and earnings (Cancio, Silva, Evans, & Maume, 1997; K. Gray, Huang, & Jie, 1993; Groot, Oosterbeck, & Stern, 1995; Grubb, 1995b, 1996, 1997; 1998, August; Hollenbeck, 1993; T. Kane & Rouse, 1993, 1995a, 1995b; Knox et al., 1993; Leigh & Gill, 1997; Rivera-Batiz, 1998; Surette, 1997).  The data sets employed in these analyses have been: the National Longitudinal Study of the High School Class of 1972 (1986 follow-up); the National Longitudinal Study of Youth (1976 through 1983 high school graduates followed up in 1989, 1990, and 1993); the 1992 National Survey of Adult Literacy; the 1985 wave of the Panel Study of Income Dynamics; and the 1984, 1987, and 1990 cohorts from the cross-sectional Survey of Income and Program Participation, which includes individuals between the ages of 25 and 64.  In these analyses, statistical controls were made for important confounding variables such as race, socioeconomic background, secondary school grades, ability (as measured by standardized test scores), age, job experience, job training, marital status, and the like (depending on the data set analyzed).  Taking the results from these published and unpublished studies, we estimate that the average net annual earnings premium for a bachelor’s degree (versus a high school diploma) is about 37% for men and about 39% for women.  The hourly wage premium is about 28% for men and about 35% for women.[5], [6]  Such average estimates fall at the upper end of our 1991 estimates of a net earnings premium for a bachelor’s degree of between 20-40%.  This finding perhaps reflects the increase in the size of the earnings premium for a bachelor’s degree in the 1980s and 1990s.[7]

            One of the major contributions of the literature of the 1990s has been its concern, not only with the economic payoff of obtaining a bachelor’s degree from a four-year institution, but also with estimating the net earnings premium for different levels of sub-baccalaureate education.  The focus of this concern has been primarily on the payoff to an associate’s degree from a community college, but attention has also been paid to the returns to vocational certificates and to postsecondary credits or vocational training completed without a degree or certificate.  The research was largely silent with respect to the economic returns to sub-baccalaureate education in our previous synthesis.  Not surprisingly, much of the important evidence in this area was uncovered in the same studies, cited above, that estimated the net premium to a bachelor’s degree with nationally-representative samples (Groot et al., 1995; Grubb, 1995a, 1996, 1997, 1998, August; Hollenbeck, 1993; T. Kane & Rouse, 1993, 1995a, 1995b; Leigh & Gill, 1997; Rivera-Batiz, 1998; Surette, 1997).  Additional evidence on the net premium to sub-baccalaureate education is provided in investigations by Grubb (1992a), Kerckhoff and Bell (1998), and Lin and Vogt (1996).  Evidence yielded by the total body of studies comes from analyses of: the 1986 follow-up of 22 to 24 year olds in the 1980 High School and Beyond sample; the 1992 National Survey of Adult Literacy; the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972; the National Longitudinal Study of Youth (1976 through 1983 high school graduates followed up in 1989, 1990, and 1993); and the 1984, 1987, and 1990 cohorts of individuals 25-64 years of age from the cross-sectional Survey of Income and Program Participation.  Depending on the individual study, statistical controls were introduced for such factors as: race, socioeconomic origins, secondary school grades and program type, ability (as measured by standardized test scores), mental status, age, job experience, job training, and the like.

            With a few exceptions, the majority of the estimates of the net economic premium attributable to an associate’s degree were statistically significant.  Aggregating the evidence across all of the above studies, we estimate that the average net annual earnings premium for an associate’s degree (compared to a high school diploma) is about 17.5% for men and about 27% for women.  The hourly wage premium is about 13% for men and 22% for women.[8], [9]  These estimates are somewhat smaller than the typical earnings premium for an associate’s degree, unadjusted for confounding influences.  For example, Grubb (1996) provides the mean annual earnings for individuals age 25-64 in the years 1984, 1987, and 1990 of the Survey of Income and Program Participation; and corresponding earnings figures for the years 1995 and 1996 are provided by the Current Population Survey (Is college still worth the cost?  The private investment value of higher education 1967 to 1996, 1998, March).  Across all five years, men with an associate’s degree had an annual earnings advantage of 27% over men with a high school degree, while the corresponding advantage for women with an associate’s degree was 40%.  Still, the net economic returns to an associate’s degree from a community or two-year college represent substantial earnings advantages over a high school diploma for both men and women (Grubb, 1998, August; Paulsen, 1998).  Furthermore, as suggested by Leigh and Gill’s (1997) analyses of the National Longitudinal Study of Youth data through the 1993 wave respondents, the positive returns to an associate’s degree are essentially the same size for experienced adult workers who return to school as they are for continuing high school graduates.

            There is also a small body of evidence that estimates the economic returns to postsecondary certificate programs.  Certificate programs, as described by Grubb (1998, August), are typically one year in length and focus on occupational rather than academic preparation or general education.  The certificate is a common credential in vocational and proprietary schools.  While the weight of evidence suggests that they can increase earning power, particularly for women, the average net economic returns to such certificates (compared to a high school diploma) appear to be somewhat less certain, and probably smaller, than the average net returns to associate’s degrees.  As reviewed by Grubb (1998, August), analyses of the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972 (Hollenbeck, 1993) and the 1992 National Adult Literacy Survey (Rivera-Batiz, 1998) found only small and statistically nonsignificant effects of certificates for both men and women.  On the other hand, Grubb’s (1997) analyses of the 1984, 1987, and 1990 cohorts of the Survey of Income and Program Participation found statistically significant, positive earnings effects of certificates for women across all three years.  The effects for men were positive and statistically significant in 1984 and 1987, but small and nonsignificant in 1990.  Results generally consistent with those of Grubb (1997) are also reported by Kerckhoff and Bell (1998) and Surette (1997).  Analyzing the 1986 follow-up of the 1980 High School and Beyond data, Kerckhoff and Bell found a statistically significant positive effect of a vocational license-certificate on the hourly wages of women, but not on the wages of men.  Surette’s analyses of men only in the National Longitudinal Study of Youth found a small, but statistically significant effect for the completion of vocational training.  However, as Grubb (1998, August) points out, this might not be the same as completing a certificate.

            Finally, the 1990s have seen a concern with estimating the net economic premium (compared to a high school diploma) of having different amounts of postsecondary education without completing a degree or credential (e.g., Grubb, 1995a, 1997; Hollenbeck, 1993; T. Kane & Rouse, 1995b; Knox et al., 1993; Leigh & Gill, 1997; D. Lewis et al., 1993; Rivera-Batiz, 1998; Surette, 1997).  As previously described, each of these investigations analyzes national samples, and introduces statistical controls for salient confounding influences.  This research has been concisely reviewed by several scholars (Boesel & Fredland, 1999; Grubb, 1998, August; Paulsen, 1998).  Their syntheses of the evidence would suggest the following generalizations.  First, individuals can potentially increase their earnings in the labor market by obtaining modest amounts of postsecondary education or vocational training without obtaining a degree or certificate.  However, the average economic premium appears to be less certain and smaller in magnitude than the average economic premium yielded by completing an associate’s degree or a vocational certificate.  Second, the size of the premium depends substantially on what subject matter one takes.  (As we shall see in the subsequent section of this chapter on within-college effects, this second point also holds for a bachelor’s and an associate’s degree.)  Third, a year of full-time enrollment can lead to a net increase in earnings over a high school diploma of about 5% or more; and the payoff of completing a year of academic credits at a community college appears to be at least equal to, if not larger than, the payoff of completing the same number of credits at a four-year college.  Fourth, and finally, while there appears to be a statistically significant return to taking a year’s worth of credits at a community college, it is unclear that any real benefit is derived from taking small numbers of community college credits (e.g., one or two courses).[10]

Credential or Program Effects

            In our 1991 synthesis, we uncovered a small body of evidence suggesting that one receives an earnings “bonus” for completing the bachelor’s degree above and beyond the economic return for having the equivalent of four years of college (e.g., 120 credits) but not completing a bachelor’s degree.  The economic literature often refers to this additional earnings increment associated with completing a degree as a “sheepskin effect” or a “credentialing effect” (e.g., Arkes, 1999; Belman & Heywood, 1991; Jaeger & Page, 1996).  Others (e.g., Grubb, 1997, 1998, August) use the term “program effect” to indicate that a degree represents a coherent sequence of courses in a field of study or discipline, as well as a program of general education.  Regardless of the descriptive term employed, the research of the 1990s not only presents substantially more evidence concerning the credential or program effect attributable to obtaining a bachelor’s degree, it also estimates the corresponding credential/program effect linked to sub-baccalaureate degrees and certificates.

            Estimating the credential/program effect of different postsecondary degrees and certificates has been largely the concern of economists (Arkes, 1999; Belman & Heywood, 1991; Frazis, 1993; Grubb, 1996, 1997, 1998, August; Heywood, 1994; Jaeger & Page, 1996; T. Kane & Rouse, 1995a; Surette, 1997).  This body of studies analyzed data from a range of nationally-representative samples.  These include: the 1979 and 1986 follow-ups of the National Longitudinal Study of the High School Class of 1972; the 1989, 1990, and1993 follow ups of the National Longitudinal Study of Youth; the 1984, 1987, and 1990 cohorts of the National Survey of Income and Program Participation; and various iterations of the Current Population Survey.  The typical analytical approach was to regress either hourly wages or annual earnings on a prediction model that specified highest degree or certificate obtained, number of years of postsecondary education completed if no degree was obtained, and, depending upon the specific study, statistical controls for important confounding influences (e.g., tested ability, labor market experience, socioeconomic background, and the like).

            Consistent with the conclusion from our 1991 synthesis, the weight of evidence from this research suggests that the individual who completes a bachelor’s degree obtains a statistically significant earnings advantage over a similar individual with the equivalent of four years of college credits, but no degree.  The magnitude of this earnings advantage is more difficult to determine.  However, across all studies our best estimate is that men with a bachelor’s degree earn, on average, about 15% more than men with four years of college credits but no degree.  For women, the corresponding earnings advantage is about 12%.  Although the estimates are quite variable and not as consistent as those for the bachelor’s degree, the weight of evidence would also suggest the presence of statistically significant credential/program effects for the associate’s degree.  Combining the results from all studies using national samples that provide relevant evidence (Arkes, 1999; Grubb, 1997; Jaeger & Page, 1996; T. Kane & Rouse, 1995a; Surette, 1997), we estimate that men who finish an associate’s degree earn, on average, about 9% more than men with the equivalent of two years of postsecondary education, but no degree.  For women, the corresponding earnings advantage for completing an associate’s degree is about 11%.[11]  We would caution, however, that these estimates, as well as those for a bachelor’s degree, are somewhat rough and may not be particularly robust.

            Finally, although it is not unequivocal, there is also a modicum of evidence from nationally-representative samples to suggest a credential/program effect for completion of vocational training.  For example, Grubb’s (1997) analyses of the 1984, 1987, and 1990 cohorts of the National Survey of Income and Program Participation found that, across all three years, women who obtained a vocational certificate had an average earnings advantage of about 10% over women with one year of postsecondary credits but no credential.  For men, the corresponding advantage was about 10% in 1984, but decreased to near parity or a slight disadvantage in 1987 and 1990.  On the other hand, Surette’s (1997) analyses of the National Longitudinal Study of Youth through 1993 found that men who completed vocational training had a statistically significant 5% advantage in hourly wages over men with the required postsecondary credits but no degree.  Our conclusion then is that the credential/program effect of completing vocational training, while likely real, is somewhat less certain and smaller in magnitude than the credential/program effect for either the associate’s degree or the bachelor’s degree.

Private Rate of Return

            Evidence establishing the net earnings premium of postsecondary degrees provides a perspective on only one part of the economic returns picture.  Premium research focuses primarily on benefits, without considering the attendant costs.  Yet, postsecondary education often requires a financial investment on the part of the student in the form of tuition, books, and other educational fees.  Moreover, for a substantial number of students, the time they invest in postsecondary education is a time during which they forego income or, if they work part-time during college, at least part of the income that they would have earned had they entered the labor force immediately after high school.  Such foregone earnings are sometimes referred to as the opportunity costs of attending college.  Attempts to take the full range of costs into account when estimating the economic returns to postsecondary education has spawned a line of inquiry we will refer to as private (or internal) rate of return research.

            Basically, private rate of return is an attempt to estimate one’s percentage return on investment.  Not surprisingly, the actual computation can get pretty complicated and esoteric due to a number of assumptions that must be considered, such as inflation on foregone earnings (e.g., Alsalam & Conley, 1995; Becker, 1992; Cooper & Cohn, 1997; Geske, 1996; Leslie, 1990; McMahon, 1991).  However, a simple way to visualize at least the fundamental concept of private rate of return to a bachelor’s degree is to divide the difference between average posttax earnings of bachelor’s degree holders and high school graduates by the sum of the private unsubsidized costs of education, plus foregone earnings.  For illustrative purposes, consider the following example using fictitious numbers for simplicity.  Suppose that the average annual posttax earnings of all male bachelor’s degree holders in the county in 1989 was $30,000 and the corresponding average posttax earnings of male high school graduates was $22,000.  Therefore, a male with a bachelor’s degree could expect to earn on average during his working life $8,000 more per year ($30,000 - $22,000) than he would be earning with only a high school degree.  Let’s also suppose that the average total unsubsidized costs of a college education (combining private and public institutions) in 1989, plus average foregone earnings if one didn’t work during college were $60,000.  If postsecondary education were considered an investment, such an arrangement would be the equivalent of purchasing a promise to receive an average of $8,000 annually during one’s working life at a present cost of $60,000.  If we divide $8,000 by $60,000, we see that the average annual yield of investing in a bachelor’s degree is 13.3%.  This 13.3% would be considered the private rate of return to a college degree.[12]

            In our previous synthesis, we concluded that the average private rate of return to a bachelor’s degree, based on studies covering the time period from 1940 to 1982, was somewhere between 11.8 and 13.8%.  When this was adjusted for differences in intellectual ability between high school and college graduates, the private rate of return for a bachelor’s degree fell to between 9.3 and 10.9%.  The evidence we uncovered in our present synthesis suggests that this private rate of return to a bachelor’s degree has remained stable or, parallel to the earnings premium for a bachelor’s degree, perhaps even increased in the late 1980s and early 1990s (Arias & McMahon, 2001; Cohn & Hughes, 1994).  As with our literature review for How College Affects Students, we have had the benefit of a number of excellent reviews of the private rate of return findings in shaping our present synthesis (e.g., Alsalam & Conley, 1995; Becker, 1992; Boesel & Fredland, 1999; Cohn & Hughes, 1994; Geske, 1996; Leslie, 1990; McMahon, 1991; Paulsen, 1998).  Although they differ in the literature they review, all of these syntheses provide a rather consistent estimate of the average private rate of return to a bachelor’s degree at around 12%, with a typical range from about 9-16%.

            From one perspective, these estimates may be biased upward because they typically are not corrected for ability or intelligence.  However, an interesting paper by Arias and McMahon (2001) uses recent studies of identical twins to estimate the average bias to ability and measurement error at about 12%.  Applying their adjustment, our estimate of the average private rate of return to a bachelor’s degree, controlling for ability, would be about 10.6%.[13]

            From another perspective, however, the unadjusted estimates of the private rate of return to a bachelor’s degree may underestimate the true private rate of return because they do not take into account other monetary returns such as health care, retirement, stock options, and support for continuing professional development.  These and related fringe benefits tend to be more substantial in the kinds of jobs held by college graduates (Boesel & Fredland, 1999; Geske, 1996).  There is also the issue of foregone earnings.  Since so many students work while attending college, the assumption of many private rate of return estimates that students will forego all earnings while obtaining their bachelor’s degree seems untenable (Cohn & Rhine, 1989).  Indeed in analysis of the 1985 wave of the Panel Study of Income Dynamics, Cooper and Cohn (1997) found that when they took into account, along with other factors, the average earnings of a student while attending college, the private rate of return to a bachelor’s degree ranged from 12.1 to 19.3%.

            Even if one assumes that the private rate of return is what we estimate the average to be, 12%, such a rate of return compares quite favorably with other investments (Boesel & Fredland, 1999; Cooper & Cohn, 1997).  As Boesel and Fredland point out, returns on the stock market have typically averaged around 11%, but, unlike private rate of return, the stock market rates are nominal returns that disregard inflation.  Moreover, if one considers the option value of a bachelor’s degree (e.g., the option of entering graduate or professional school), as well as the nonmonetary returns (e.g., health benefits, working conditions, lifelong learning, enhanced life chances for children (see Leslie, 1990; Mathios, 1989; McMahon, 1998), a college degree continues to be a reasonably informed and prudent investment.[14][15]

Causal Mechanisms

            Although the evidence is quite clear that bachelor’s and associate’s degrees provide substantial occupational prestige and earnings premia to individuals who obtain them, it is not always as clear just why this is the case.  Determining the causal mechanism(s) underlying the positive link between postsecondary education and both occupational prestige and earnings has become one of the favorite indoor sports of both economists and sociologists.  In our 1991 synthesis, we concluded that no single causal mechanism provided a completely satisfactory explanation and that a number of processes may be at work.  We uncovered little evidence in the decade of the 1990s to suggest a fundamentally different conclusion.

            As suggested by Bills (2000), there are at least seven distinct theories or explanations that economists and sociologists have offered for why those with the most schooling get the most desirable and best jobs.  Since a detailed discussion of these theories is beyond the scope of this book, we confine our synthesis to the evidence regarding three of the major theories: human capital, signaling/screening, and credentialing.  Human capital theory suggests that college graduates have more desirable jobs and earn more than high school graduates because postsecondary education provides the former with marketable skills and abilities relevant to job performance.  Signaling/screening are two complementary mechanisms in that job seekers signal and employers screen (Bills, 2000).  Postsecondary education may not so much influence the cognitive and personal traits related to job productivity as simply select for individuals who have such traits to begin with.  Thus, a college degree can be used by job seekers to signal desirable intellectual and personal traits, irrespective of whether those traits are acquired as the result of postsecondary education.  Employers can use a college degree as a relatively inexpensive screening device to select individuals who they believe possess intellectual skills and personal traits predictive of productivity for the best jobs.  Finally, credentialism posits that employers may not select or reward individuals solely on the rational basis of potential or actual productivity.  Rather, the factors that influence these decisions are shaped by such things as social class, snobbery or, as suggested by Bills (2000, p. 20) “widely shared societal assumptions about the appropriate relationship between schooling and job assignment.”  This would mean that those with postsecondary degrees could end up being overly positioned or rewarded in the labor market for reasons unrelated to individual productivity (Jencks et al., 1979).

            While we found evidence to support each of these explanations, none seems sufficient to unambiguously account for the relationship between educational attainment and labor market rewards.  For example, the most straightforward explanation is probably human capital; and the underlying premise that postsecondary education provides skills that make individuals better employees has considerable logical appeal.  It seems almost axiomatic that a bachelor’s degree in such fields as engineering, accounting, nursing, and speech pathology, to name a few, indicates the completion of a course of study that actually provides knowledge and skills important to effective job performance.  Not inconsistent with this view are Grubb’s (1996; 1997) findings for both men and women that the economic returns to bachelor’s and associate’s degrees tend to be more pronounced when one’s academic major is closely related to one’s job than when it is unrelated.  Similar results are reported for two national samples of bachelor’s degree recipients by Tsapogas, Cahalan, and Stowe (1994) and for graduates of single institutions by Callaway, Fuller, and Schoenberger (1996), Dutt (1997), and Fuller and Schoenberger (1991).  Presumably, if the skills one acquires in his or her program of study are applicable to the job requirements, the economic returns increase, a result generally compatible with human capital theory.  However, as Grubb points out, this may also be explained by the fact that those academic majors that are linked to the highest economic returns are also the ones most likely to lead to related employment (e.g., engineering, business, health).

            Whether evidence suggesting a strong relationship between educational attainment and labor market success is simply the result of increasing one’s human capital is nearly impossible to verify in the evidence we reviewed.  For example, in their analyses of the 1986 follow up of the National Longitudinal Study of the High School Class of 1972, Knox, Lindsay, and Kolb (1993) found an almost monotonic positive relationship between amount of formal postsecondary education completed and both occupational status and earnings, even after controls were introduced for ability test scores, race, gender, and socioeconomic status.  Similar findings are reported for earnings by Arkes (1999) in analyses of the 1993 wave of the National Longitudinal Study of Youth.  One could reasonably view such findings through the lense of a human capital perspective and conclude that, net of ability, the greater one’s acquisition of high level knowledge and skills, as indicated by amount of exposure to postsecondary education, the greater one’s returns in the labor market.  On the other hand, such evidence may merely suggest that years of postsecondary education or degrees signal important personal skills or attributes that employers value because they predict job productivity.  For example, Arkes (1999) concluded that a bachelor’s degree signals intellectual ability to employers.  However, it was also the case that bachelor’s and associate’s degrees provided an earnings premium above and beyond intellectual ability and the equivalent numbers of credit hours required.  This suggests that these degrees may signal personal attributes to employers that they value as predictors of job productivity, other than ability (e.g., ambition, motivation, persistence).

            Other evidence reported by Grubb (1993) is purported to support the screening/signaling hypothesis, at least in part.  Grubb reasoned that if degrees signaled ability or other desirable traits to employers, then they would leave a stronger impact on earnings in salaried occupations, which are presumably screened, than on the earnings of those who were self-employed.  Using the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972, and controlling for such factors as ability, job experience, socioeconomic status, and high school grades, Grubb found mixed support for his hypothesis.  Vocational associate’s degrees counted more in salaried (screened) than in self-employed (unscreened) positions, while generally the reverse was true for the bachelor’s degree.  Grubb concluded that the labor market for sub-baccalaureate credentials works differently than it does for bachelor’s degrees.  Such a finding further underscores the difficulty one has in uncovering a single, or perhaps even a predominant, explanation for the education-earnings relationship.

            Evidence supporting a credentialing explanation for the fact that the more highly educated have the most desirable and best paying jobs rests largely on the evidence we reviewed in the previous section of this chapter on credential or program effects.  As we saw in that section, the weight of evidence was reasonably clear that individuals receive an earnings bonus for completing a bachelor’s or associate’s degree above and beyond the economic return of having the equivalent years of college (four or two, respectively), but not completing the degree.  It is highly questionable that the final year of postsecondary education leading to either the bachelor’s or the associate’s degree actually enhances individual productivity at a higher rate than the preceding years.  Thus, through a credentialing lense, degrees may function as socially-sanctioned gatekeepers by which those who have them gain easier access to higher-paying jobs and career paths than those who do not, for reasons not necessarily related to productivity.  Put another way, postsecondary degrees are less about conferring labor market skills, or signaling ability, than they are about conferring status that can be used in American society to gain entry into the most prestigious and rewarding occupations.

            Of course, the earnings bonus or boost associated with completion of postsecondary degrees does not necessarily lead to credentialing explanation.  Completing a degree might signal personal traits such as perseverance or focus that are important to employers because they predict job productivity.  Moreover, a degree may represent completion of a coherent, integrated program of study that is more predictive of job relevant skills than simply completing an equivalent number of postsecondary credit hours.

            What seems evident is that the causal mechanisms underlying the relationship between educational attainment and both occupational positioning and earnings are complex.  They may function differently, and with varying degrees of importance, in different career paths, at different times in one’s career, in different jobs or labor market sectors, and with changes in the economy and the nature of work.  It may be fruitless to search for a single, dominant explanation.  Furthermore, the increased importance of computers and information technology, and how they influence fundamental notions of work and career may be an additional wild card that shapes broad-based societal perceptions of competence and competitiveness in the labor market (Bassi, 1999).

A Final Word

            In this section, we have attempted to summarize the evidence on the net effects of college on career and economic returns.  Our estimates are based on the average returns that accrue to an individual, irrespective of the type of postsecondary institution attended or one’s academic and nonacademic experiences once there (e.g., major field of study, grades, extracurricular involvement, and the like).  Consequently, they potentially mask variations in between- and within-college effects.  We turn to these in the next two sections of the chapter, starting with between-college effects.

Between-College Effects

Conclusions from How College Affects Students

            The most investigated of all institutional characteristics is that of institutional “quality,” typically assessed in terms of student body selectivity (e.g., average ACT or SAT score of entering students) or reputational and prestige indexes.  Compared with other institutions, elite or selective institutions tend to enroll students with high occupational status aspirations to begin with, and their impact appears to be one of maintaining or perhaps slightly accentuating the status level or academic career orientation of initial choice.  This net impact on career choice is quite small compared to that attributable to career choice at the beginning of college.  It may be particularly true of students attending selective or prestigious institutions that the undergraduate experience is used more to implement than to choose a career.

            Attendance at a selective college modestly increases the likelihood that women will choose sex-atypical (male-dominated) majors and careers and that they will enter sex-atypical occupations.  It also appears that a degree from an elite institution confers a slight advantage in various dimensions of career mobility and success (e.g., technical or supervisor responsibility, level of managerial attainment).  However, with the possible exception that college selectivity may have more positive implications for attainment in the professions than in managerial or business occupations, the weight of evidence indicates that attending a selective or prestigious institution has little net impact on overall job status, job productivity, or job satisfaction.

            Net of other factors, college quality (and particularly selectivity) has a small positive direct effect on earnings.  The best estimate of the magnitude of this effect is that quality indexes account for between 1 and 1 1/2 percent of the variance in individual earnings above and beyond other factors.  There is some evidence that this effect is nonlinear; only those colleges at the very top of the distribution of selectivity or academic reputation may significantly enhance earnings.  Estimates of direct effects may underestimate the total positive impact of institutional quality measures on earnings.  Institutional quality may also have a positive effect on earnings by enhancing educational attainment and attendance at prestigious professional schools.  We conclude that the evidence is more supportive of a screening (as opposed to a human capital) explanation for the apparent impact of college quality on earnings.

            Comparison of two-year and four-year institutions has produced the most pronounced and consistent between-college effects on occupational status.  Net of other factors, students who begin the postsecondary education experience in two-year colleges have significantly lower job status than those who start at four-year institutions.  Most of this difference, however, appears to be attributable to the adverse impact of two-year institutions on educational attainment.  For individuals of equal educational attainment, whether they start at two-year or four-year institutions makes little difference in early occupational status, employment stability or job satisfaction.  Similarly, when individuals of equal background traits and educational attainment are compared, any direct earnings penalties for attending a two-year college are quite small early in the career, though they may increase slightly with longer work experience.  It is likely, however, that initial attendance at a two-year college may have a discernible negative indirect effect on earnings due in large measure to its inhibiting influence on educational attainment.

            There is evidence that men’s colleges have independently enhanced male career choice and attainment in such areas as business, law, and the professions in general.  Substantially more research, however, has focused on the impact of women’s (versus coeducational) institutions.  The weight of evidence suggests that attending a predominantly women’s institution rather than a coeducational one has little or no independent impact on a woman’s career salience (interest in or commitment to a career), the status or prestige level of the job she obtains, her earnings, or the likelihood of her actually entering a sex-atypical career (globally defined according to the percentage of men in the field).  On the other hand, women’s institutions appear to enhance orientation toward a sex-atypical occupation during college, entrance in certain specific sex-atypical occupations (such as medicine and scientific research), and prominence or achievement within a specific occupational status level.

            Net of other factors, attending a predominantly black institution rather than a predominantly white institution appears to have only a trivial impact on the occupational status of black men or black students generally.  However, some evidence suggests that attendance at a black college may enhance the early job status of black women.  There was little consistent evidence to suggest that college racial composition had a statistically significant net impact on the earnings of black men or women.

            Attending a large institution appears to have a small positive influence on occupational status and earnings that is independent of student background characteristics and the selectivity of the student body.  There is parallel evidence to suggest that major research universities, most of which are large, also positively influence earnings; but it is difficult to separate this effect from institutional quality.

            Institutional control appears to have little consistent impact on career choice, occupational status, or women’s entry into sex-atypical careers.  However, public control appears to enhance the likelihood of successfully implementing career plans for becoming an engineer or college teacher while reducing the likelihood of successfully implementing plans for law, business, medicine or nursing.  The major influence of liberal arts colleges may be in their enhancing of women’s choice of sex-atypical majors and careers, although the evidence supporting this conclusion is not particularly strong.  Net of other factors, attending a liberal arts college would appear to have little or no impact on occupational status.

            The most consistent college environmental impact on career choice appears to be that of “progressive conformity.”  Progressive conformity hypothesizes that student career choice will be influenced in the direction of the dominant peer groups in an institution.  A small amount of evidence indicates that irrespective of initial career choice, seniors tend to be planning careers consistent with the most typical academic majors in their institution.  There is also evidence, though less of it, to suggest that independent of initial career choice, a student’s likelihood of actually working in a particular occupation increases with the percentage of majors at his or her college corresponding to that occupation.

            There is modest support for the expectation that transfer between four-year institutions has negative consequences for both early career occupational status and earnings.  Most of this negative effect is indirect through the inhibiting influence of transfer on educational attainment.

            Considering only four-year institutions, the weight of evidence suggests that any statistically significant between-college effects are quite modest in magnitude.  This is particularly the case when compared to the general net effects of attending rather than not attending college.

Evidence from the 1990s

            We uncovered a substantial body of studies conducted during the decade of the 1990s that focused on between-college impacts on career and economic returns.  Much of this research is uneven in terms of methodological rigor; and perhaps in part because of these methodological problems it is difficult to find evidence consistent enough to permit unequivocal conclusions.  Even in those areas where the evidence is relatively strong (e.g., the impact of college selectivity on earnings), there are alternative findings or explanations which tend to muddy the waters.  The same fundamental methodological problem that accompanies any estimate of between-college effects is particularly relevant in determining the between-college effects on career.  Specifically, there is great variability in the cognitive abilities, socioeconomic backgrounds, career aspirations, and ambitions of students attending different kinds of postsecondary institutions (e.g., Behrman, Kletzer, McPherson, & Schapiro, 1995; Behrman, Rosenzweig, & Taubman, 1994; L. Lewis & Kingston, 1989; Lillard & Gerner, 1999; Sazama, 1994).  Furthermore, such individual student characteristics are likely to play a major role in different dimensions of career choice and success.  For example, a number of economists have noted that the economic returns to cognitive skills (i.e., the correlation between scores on standardized cognitive tests and earnings) has increased over the past several decades (e.g., Hoxby & Long, 1999; Murnane, Willett, Duhaldeborde, & Tyler, 1998; Murnane, Willett, & Levy, 1995; Neal & Johnson, 1994).  Similarly, one of the strongest predictors of eventual occupational attainment (e.g., occupational status) is occupational aspirations or ambition when entering college (e.g., Inoue & Ethington, 1997; Kingston & Smart, 1990; Stoecker & Pascarella, 1991; Whitaker & Pascarella, 1994).  Consequently, the relationship between the type of college attended (e.g., selective versus nonselective) and any particular career outcome (e.g., earnings) is likely to be substantially confounded by differences in the career-salient characteristics of the students who attend different kinds of colleges.

            Estimating between-college effects on career and economic returns is also complicated by other factors.  For example, postsecondary institutions not only differ dramatically in the kinds of students they recruit and enroll, they may also differ dramatically in what it costs the individual to attend them (Choy, 1999).  For example, Morganthau and Nayar (1996) point out that the average cost of attending an elite private college (e.g., an Ivy League, or similar, school) is about $1,000/week, while the average cost of attending a public university is about one-fourth of that, or about $250/week.  Moreover, it is clearly the case that many measures of institutional “quality” (e.g., selectivity, academic reputation, prestige) are confounded by whether or not the institution is private.  Even though the “real” costs of attending college may be less because of widespread financial aid in the form of student aid, grants, fellowships, tuition waivers, and the like, these differences in costs are still substantial, and undoubtedly need to be taken into account when estimating the earnings’ returns accruing to the graduates of different kinds of colleges.

            A second factor is major field of study.  As we will see in the section on within-college effects in this chapter, a student’s major field of study is, unsurprisingly, a major determinant of one’s eventual occupation and earnings.  However, it is evident that different types of colleges offer their students different kinds of academic majors.  For example, as pointed out by Jacobs (1999), more selective/prestigious, private institutions tend to focus on academic fields of study that lead to more lucrative jobs (e.g., engineering, business, science).  Conversely, because of their state-oriented mission, less selective, public institutions may be expected, if not required, to offer academic majors that lead to less lucrative occupational paths (e.g., education, social work, home economics).  Thus, if academic field of study is not taken into account, it may be easy to attribute the earnings or occupational status differences of graduates to an institutional effect when it is really the result of one’s major field of study.

            Finally, the estimation of between-college effects on career and economic returns is also complicated by the fact that a substantial number of students in the American postsecondary system attend more than one college or university before earning their bachelor’s degree (Adelman, 1998a).  However, as opposed to such outcomes as learning, cognitive development, and values and attitudes, between-college impacts on career and economic returns do not necessarily assume a human capital or socialization influence (i.e., that some colleges provide a higher quality education than others).  Rather, the impact of where one attends college may, in fact, simply reflect the extent to which completing a bachelor’s degree from that particular institution signals personal traits that employers value as predictive of job performance or productivity, irrespective of where they were acquired (e.g., high intelligence, ambition, social skills, and the like).

            The above considerations clearly make the estimation of between-college effects on career and economic returns complex and fraught with ambiguities.  Nevertheless, we have attempted to synthesize this body of evidence within the following general categories: institutional quality, institutional control, Carnegie classification, institutional size, institutional racial composition, institutional gender composition, two-year versus four-year colleges, and impact of peers.

Institutional Quality

            Research estimating the net impacts of institutional quality constituted at least 50% of the total body evidence we uncovered pertaining to between-college effects on career and economic returns.  Not surprisingly, different studies operationally defined institutional quality in different ways.  Included were such dimensions as: academic expenditures/student, faculty/student ratio, percentage of faculty with Ph.D.s, tuition costs, reputational ratings, average faculty salaries, and selectivity (typically based on the average ACT or SAT scores of entering freshmen).  An obvious problem, of course, is that all these various dimensions of institutional quality tend to be substantially and positively intercorrelated.  For example, the most academically selective institutions tend also to have the highest reputational ratings, the highest faculty salaries, the highest expenditures/student, and, because they also tend to be private, the highest tuition costs.  As a result, and because of the vagaries of multiple regression procedures when the predictors are highly correlated, determining which quality dimensions are having the strongest impact is frequently problematic.  Some researchers have dealt with this problem by creating a composite measure of institutional quality that combines several of the dimensions indicated above.  Most, however, have employed institutional selectivity (e.g., the average ACT or SAT score of incoming students) as a single proxy measure for institutional quality.  The research on the impact of institutional quality has focused on career choice, occupational status, career mobility and success, and earnings.
Career Choice

            What little evidence we uncovered suggests that institutional quality measures have only a mixed impact on students’ career choices during college.  For example, Cole, Barber, Bolyard, and Linders (1999) focused on the career choices of high achieving arts and sciences majors (grade point averages of 2.8 or above) at 34 institutions: 8 Ivy League schools, 13 liberal arts colleges, 9 large state universities, and 4 historically black colleges.  Statistical controls were introduced for an extensive set of potential confounding influences such as: race, specific freshman career interest, academic ability, college grades, interaction with faculty, influence of work experience, and the like.  In the presence of such controls, attending an Ivy League institution (versus all others) had no significant impact on choosing law, medicine, or college teaching as a career.  Attending an Ivy League school did modestly, but significantly increase the likelihood that one would choose business as a career.  However, this increase was essentially attributable to differences in senior-year business career choice between students at Ivy League schools and seniors at large public universities.  In all four career choices (law, medicine, business, and college teaching), initial interest in a career as a freshman was, by far, the strongest predictor of senior career choice.

            Tusin (1991) analyzed the 1971-80 Cooperative Institutional Research Program data to determine why women choose elementary and secondary school teaching as a career.  Net of a battery of potential confounding influences, including freshman-year career choice, institutional selectivity had a modest, but statistically significant negative influence on choosing elementary or secondary school teaching as a career.  Such a finding may reflect the influence of faculty and peer cultures at selective institutions in shaping a student’s career aspirations and choice.  The normative press of the culture at selective, elite institutions may function to steer student aspirations toward career choices that are perceived as more lucrative and “prestigious” than teaching in elementary or secondary schools.  At the same time, the effect could just as easily be attributable to the fact that more selective colleges and universities, particularly if they are private, are substantially less likely to offer education and teacher preparation as a major field of study.

Occupational Status

            Consistent with conclusions from our 1991 synthesis, we found little evidence to suggest that measures of institutional quality have more than a trivial and statistically nonsignificant direct impact on overall occupational status.  Analyzing the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972, Knox, Lindsay, and Kolb (1993) introduced statistical controls for such factors as tested academic ability, race, gender, socioeconomic background, college grades, major field of study, educational attainment, and the like.  In the presence of these controls, the selectivity of the institution attended had a small and nonsignificant effect on the occupational status of the job one held in 1986.  Remarkably consistent results are reported by Avalos (1996), analyzing the 1994 follow-up of the Cooperative Institutional Research Program’s 1985 Freshman Survey, and by Dey, Wimsatt, Rhee, and Waterson (1998), analyzing the 1974-75 and 1992-93 follow-up of 1,957 high school seniors from the Wisconsin Longitudinal Study.  Both studies employed an analytical design similar to that of Knox et al. and introduced statistical controls for salient confounding influences.  In the Avalos study, institutional selectivity failed to have a significant direct impact on 1994 occupational status; and in the Dey et al. study, neither institutional prestige (e.g., composite of selectivity, percent of students seeking a Ph.D., median high school grades of entering students, and ratio of high ability applicants to total number of admitted students) nor institutional resources (e.g., average faculty salary, faculty with Ph.D.s, number of library volumes) significantly influenced 1974-75 occupational status or 1992-93 occupational status.[16]

            All three of the studies cited above (i.e., Avalos, 1996; Dey et al., 1998; Knox et al., 1993) focus on estimating the net direct influence of measures of institutional quality on overall occupational status.  While the clear weight of evidence suggests that this direct influence is trivial and not statistically significant, it is likely that institutional quality may, nonetheless, have at least a modest positive, indirect effect on occupational status.  This indirect influence is attributable to the fact that (as we saw in Chapter    ) dimensions of institutional quality such as student body selectivity positively influence educational attainment, which, in turn, is a strong determinant of the prestige of the job one holds.  Unfortunately, the analytical models in the investigations by Avalos (1996), Dey et al. (1998), and Knox et al. (1993) do not permit us to estimate the magnitude or statistical significance of this indirect effect.

            The failure of college quality measures to directly influence overall job status is consistent with our 1991 conclusions.  However, also consistent with our 1991 conclusions is evidence to suggest that attending a selective college enhances occupational attainment in specific professions such as medicine and law.  For example, Lentz and Laband (1989) found that even with controls for college grades, college courses taken, Medical College Admissions Test scores, race, and parental occupation and education, the academic selectivity of the college one attended had a statistically significant, positive influence on admission to medical school.  Similarly, Kingston and Smart’s (1990) analyses of the 1980 follow-up of the 1971 Cooperative Institutional Research Program freshman survey found that attending one of the 74 most selective private colleges in the United States significantly increased one’s likelihood of completing a high status professional degree (i.e., MD, JD, MBA).  This effect persisted even in the presence of statistical controls for such factors as race, sex, family background, high school achievement, precollege occupational aspirations, and self-estimates of academic ability and drive to achieve.  Interestingly, in both studies this effect was nonlinear, and generally accrued only to those students attending the most selective or elite institutions in the country.  For Kingston and Smart, it was institutions having incoming freshmen with an average combined SAT score of 1175 or higher, while for Lentz and LaBand, it was institutions where more than 75% of the freshmen were in the top 10% of their high school class and scored over 1250 on the combined SAT.  Such institutions, at most, educated about 1 or 2% of all four-year college students in the national postsecondary system.  For the remaining 98% or so of all four-year college students, the selectivity of the institution they attended made little or no difference.

Career Mobility and Success

            Consistent with the conclusions of our 1991 synthesis, we found a small body of evidence to suggest that attending a selective college confers a modest advantage in job attainment and career mobility.  The evidence, however, is somewhat complex and suggests that college quality may signal an individual’s ability to employers, rather than conferring unique skills that make for better job performance.  Data from graduates of accounting programs in 82 universities was analyzed by Colarelli, Dean, and Konstans (1991) to determine if institutional characteristics influenced job offers and early job productivity.  Measures of institutional quality such as student body selectivity and institutional resources were both significantly and positively related to the number of job offers an individual received from the eight largest and most prestigious accounting/consulting firms in the region.  However, after one year on the job supervisors’ ratings of job performance and promotability were unrelated to institutional resources and actually had a significant negative association with institutional selectivity.

            A more focused set of longitudinal studies by Spilerman and colleagues (Ishida, Spilerman, & Su, 1997; Spilerman & Lunde, 1991) investigated the educational factors that influenced job promotion prospects in a single large insurance company.  Spilerman and Lunde (1991) introduced statistical controls for years of education, race, gender, age, seniority, and salary grade level and found that a measure of college selectivity had modest, but statistically significant positive effects on promotion in the middle organizational ranks where college training would provide relevant job skills.  The selectivity of the institution one attended had only a chance impact on the likelihood of being promoted at either the lowest or highest organizational ranks in the company.  Generally consistent results were also reported by Ishida, Spilerman, and Su (1997) in what appears to be a further study of promotion in the same company.  With controls for level of formal education, college major, age, race, sex, and seniority, institutional selectivity once again had a modest, but statistically significant positive impact on promotion to the middle organizational ranks of the company (i.e., senior management), but essentially only a chance effect on promotion at either lower or higher ranks (i.e., administrative or vice-presidential grades).

            Although it is more prominent in the Spilerman and Lunde (1991) study than in the Ishida, Spilerman, and Su (1997) study, both investigations provide evidence to suggest that at the lower and middle ranks of the firm they studied, the impact of college selectivity on promotion varied with experience or seniority in the firm.  The positive effect of college selectivity was greatest for employees who were recent or initial hires in the firm and at the early stages in their careers.  As seniority in the firm increased, and direct measures of job performance became available, the selectivity of the college one attended decreased in importance.  Spilerman and his colleagues conclude from such evidence that given lack of direct information on job performance of new hires at the beginning of their careers, the firm’s employers use college selectivity as a proxy or “signal” for the possession of intellectual and related skills that are important for job performance.  However, if college selectivity signals higher intellectual or other skills related to effective job performance, it is not clear from either the Spilerman and Lunde or Ishida, Spilerman, and Su studies if the individual acquires them from his or her experience in college or essentially entered college with them.[17]

            Finally, there is also evidence of an indirect nature that speaks to the effect of college quality on career mobility.  Robst (1995) analyzed data from 560 male heads of household between 18 and 64 years of age in the 1976, 1978, and 1985 waves of the Panel Study of Income Dynamics.  His purpose was to estimate the net impact of college selectivity on the probability that an individual was employed in a job for which he was overeducated (i.e., held a job in which his education was substantially higher than that typically required).  With controls for years of education, work experience, number of years in one’s current job, and scores on a 13-question sentence completion test, three institutional quality measures (i.e., average ACT/SAT scores of the entering freshmen, educational and general expenditures per student, and a prestige rating) had modest, but statistically significant negative effects on the probability of being overeducated.  Moreover, college selectivity was also positively associated with the likelihood of moving from being overeducated for one’s job in 1976 to being in a job in 1985 for which one was not overeducated.  The findings of the Robst study, however, are likely confounded by the inability to control for men’s precollege levels of career or occupational aspirations—strong predictors of both a man’s eventual occupational level and the type of college he attends (Pascarella & Terenzini, 1991).  Robst also candidly points out that it is questionable that a 13-item sentence completion test is an adequate measure of individual cognitive ability.  In short, the presumed negative effects of college quality on overeducation may in fact be attributable more to the characteristics of the men who attend high quality colleges.[18]

Earnings

            Perhaps the largest single body of research on between-college effects on career and economic benefits concerns the impact of undergraduate institutional quality measures on individual earnings.  Once again, institutional quality is operationally defined in different ways in different investigations, but student body selectivity appears to be the most common proxy for college quality.  The typical study in this body of research analyzes data from a nationally-representative sample, uses the natural logarithm of earnings as the dependent variable in order to adjust for positive skewness in the distribution of earnings, and introduces statistical controls for factors that potentially confound the relationship between the quality of the institution attended and an individual’s earnings.  These confounding factors include such variables as ability test scores, family socioeconomic background, race, sex, major field of study, educational attainment, and the like.  On average, the studies we reviewed explained considerably less than half of the variance (R2) in individual earnings, typically in the neighborhood of 25-35%.  Unless one is willing to accept the view that two-thirds or more of the earnings differences among college graduates are attributable to luck, it seems reasonable to conclude that a number of important influences on earnings are not taken into account in the literature we reviewed.

            This large percentage of unexplained variance does not necessarily mean that it is impossible to get a reasonably accurate estimate of the net effects of institutional quality on earnings.  However, in analyzing the evidence we would argue that, in addition to such factors as race, sex, family socioeconomic factors, educational attainment, measures of labor market experience, and the like, obtaining an unbiased estimate of the net, direct impact of institutional quality on individual earnings means that four additional influences need to be taken into account.  These are: 1) cognitive or intellectual ability, 2) ambition, 3) major field of study during college, and 4) the differential costs of attending different kinds of institutions.  As previously pointed out in this chapter, cognitive ability and ambition are important considerations because they are not only highly correlated with attending a selective or elite institution (e.g., A. Astin, 1993b; Dale & Krueger, 1999; Lillard & Gerner, 1999; Pascarella & Terenzini, 1991), but are also salient predictors of earning potential (Monks, 2000; Murnane et al., 1995; Sweetman, 1994a, 1994b; Whitaker & Pascarella, 1994).  Similarly, it is important to take into account major field of study because selective/prestigious institutions tend to offer academic fields of study that lead to the most lucrative jobs (Jacobs, 1999).  Finally, failure to account for the substantially higher costs typically associated with attending a selective/prestigious (and often private) college can lead to inflated estimates of the actual net earnings benefits associated with attendance and graduation from such institutions (Behrman, Rosenzweig, & Taubman, 1996; Brewer, Eide, & Ehrenberg, 1999; S Thomas, 1998; S. Thomas, 2000).  Nearly all of the effects of college quality on earnings is derived from secondary analyses of preexisting data sets.  The variables represented in most of these data sets simply do not permit one to introduce controls for all, or even most, of the important confounding influences.  Consequently, as suggested by Kane (T. Kane, 1998, p. 432) in a summary caution about research on college quality and earnings, “what looks like an effect of attending an elite college may really be an effect of unmeasured preexisting differences in academic or earning potential.”

            Our present synthesis is based on evidence from 26 individual published and unpublished studies which appeared between 1989 and 2000.  These investigations analyze data from numerous independent data sets.  The specific data sets and the studies that employ them were as follows:

1.     The National Longitudinal Study of the High School Class of 1972-1979, and 1986 follow-ups (Arcidiacono, 1998; Brewer et al., 1999; Dale & Krueger, 1999; Hoxby & Long, 1999; James & Alsalam, 1993; James, Alsalam, Conaty, & To, 1989; Knox et al., 1993; Loury, 1997; Sweetman, 1994a, 1994b).

2.     The High School and Beyond 1980 and 1982 cohorts followed up in 1986 and 1991, respectively (Brewer & Ehrenberg, 1996; Brewer, Eide, & Ehrenberg, 1996; Fitzgerald, 2000; Fox, 1993; Hilmer, 2000; T. Kane, 1998; Loury, 1997).

3.     The National Longitudinal Survey of Youth, 1987-89, 1993, and 1995 follow-ups (Daniel et al., 1996a; Daniel, Black, & Smith, 1996b; Hoxby & Long, 1999; Monks, 2000).

4.     The College and Beyond 1976 cohort followed up in 1995 (Bowen & Bok, 1998; Dale & Krueger, 1999).

5.     The Baccalaureate and Beyond Study of 1992-1993 graduates followed up one year later (S Thomas, 1998) and four years later (S. Thomas, 2000).

6.     The National Center for Education Statistics Surveys of Recent College Graduates: 1985-86 graduates followed up in 1987 (Rumberger & Thomas, 1993); 1989-90 graduates followed up in 1991 (Tsapogas et al., 1994).

7.     The National Science Foundation New Entrants Survey of 1992 graduates followed up in 1993 (Tsapogas et al., 1994).

8.     The Occupational Changes in a Generation, 1972 data (Hoxby & Long, 1999).

9.     The Cooperative Institutional Research Program data: 1972 freshmen followed up in 1980 (Kingston & Smart, 1990); 1985 freshmen followed up in 1994 (Avalos, 1996).

10.    The Panel Study of Income Dynamics from 1975 to 1992 (Turner, 1999, April).

11.    A survey of identical and nonidentical female twins born in Minnesota and followed-up in 1993 at about age 45-46 (Behrman, Rosenzweig et al., 1996).

            The body of evidence yielded by these investigations would suggest the following general conclusions.  First, although there are some clear exceptions (Arcidiacono, 1998; Avalos, 1996; James & Alsalam, 1993; Knox et al., 1993; S Thomas, 1998; Tsapogas et al., 1994), the weight of evidence suggests that measures of institutional quality, and particularly student body selectivity, have statistically significant, positive net impacts on subsequent earnings.  Our best estimate is that, net of other influences (including both individual student characteristics and other institutional characteristics such as private control and size), attending a college with a 100 point higher average SAT score (or ACT equivalent) is associated with about 2 to 4% higher earnings in later life. [We note that this estimate is somewhat more conservative than other summaries (e.g., Dale & Krueger, 1999; Hilmer, 2000), but this difference is likely attributable to the fact that we derive our estimates from a somewhat broader range of studies.]  Moreover, when differential tuition costs are taken into account to adjust for the fact that the most selective institutions are typically private, the positive effect of attending a selective or elite institution on subsequent earnings is reduced but does not disappear (e.g., Behrman et al., 1994; Brewer & Ehrenberg, 1996; Brewer et al., 1999; Sweetman, 1994a).  Second, consistent with the conclusions from our 1991 synthesis, there is also evidence to suggest that the impact of institutional selectivity on earnings is nonlinear.  Only those elite institutions at the very top of the selectivity distribution may have a substantial impact on earnings (e.g., Fox, 1993; Hilmer, 2000; Kingston & Smart, 1990).  Third, there is empirical support for the contention that the net impact of institutional selectivity or similar quality measures on earnings has increased over time.  Investigations that consider the effects of college selectivity for different national cohorts in different time periods tend to find that its estimated impact on earnings is of a somewhat larger magnitude in more recent than in older cohorts (Brewer et al., 1996; Hoxby & Long, 1999; Loury, 1997; Turner, 1999, April).  Finally, there is also evidence suggesting that, in addition to its statistically significant direct effect on earnings, college selectivity may also have a positive indirect effect due to its enhancement of educational attainment and graduate or professional school attendance (Arcidiacono, 1998; Eide, Brewer, & Ehrenberg, 1998).  Since these studies do not control for either precollege educational or occupational aspirations, however, it is likely that this indirect effect is biased upward by some unknown amount.

            Although most of the evidence on institutional quality and earnings employs various measures of selectivity as the primary quality indicator, there is also evidence to suggest that other quality indicators may be linked with earnings.  Unfortunately, there is little in the way of evidence that is consistent across studies analyzing different samples.  For example, with controls for other college characteristics as well as individual-level confounding variables, Daniel, Black, and Smith (1996a; 1996b) found that expenditures per student had a significant, positive effect on wages for men, though not for women.  However, there is only mixed support for this finding in the work of Behrman, Rosenzweig, and Taubman (1996), Dale and Krueger (1999), and Fitzgerald (2000), and none at all in findings reported by James and Alsalam (1993) and Tsapogas, Cahalan, and Stowe (1994).  Similarly, Tsapogas, Cahalan, and Stowe found that percent of faculty with a Ph.D. positively influenced earnings in one national sample they analyzed, but not in the other.  Moreover, there was little support for the unique, positive impact of percent of faculty with a Ph.D. in the earnings functions of Daniel, Black, and Smith (1996a; 1996b). Behrman, Rosenzweig, and Taubman did find that average faculty salaries at the institution attended positively influenced the subsequent earnings of women, but we uncovered no independent replication of their evidence.

            Aside from various measures of institutional selectivity, we uncovered only one institutional quality indicator, faculty/student ratio, that was found to have a significant, positive net effect on earnings across independent samples.  Both Behrman, Rosenzweig, and Taubman (1996) and Daniel, Black, and Smith (1996b) found that attending an institution with a high faculty/student ratio had a significant, positive effect on earnings, net of other factors.  However, even here the overall findings are inconsistent.  In their analyses of two independent national samples, Tsapogas, Cahalan, and Stowe (1994) found that an institution’s faculty/student ratio had a significant, net positive effect on earnings in one sample, and a significant, net negative effect on earnings in the other sample.  Similarly, Fitzgerald's (2000) analyses of the 1991 follow-up of the 1980 High School and Beyond cohort reported that an institution's ratio of faculty to students had no net impact on women's earnings and a small negative effect on the earnings of men.

            Thus, the bottom line would appear to be that, when institutional quality is defined largely in terms of academic or student-body selectivity, it has a generally consistent positive effect on subsequent earnings.  Our estimate is that each hundred point increase in the average SAT score (or ACT equivalent) of the entering students at a college increases earnings by about 2 to 4%, though earnings may be most clearly enhanced by attending an institution at the very highest or elite levels of the selectivity distribution.  We would argue, however, that the body of research evidence on which we base this conclusion probably provides an inflated estimate of the impact on subsequent earnings of having a bachelor’s degree from a selective institution.  Due in large measure to the fact that they are generally conducting secondary analyses of existing data sets, nearly every investigation we reviewed in this body of research was unable to control for one or more salient confounding variables.  This was the case for even the most meticulously conducted and methodologically rigorous studies.  For example, Brewer and Ehrenberg (1996) considered cognitive ability and differential tuition costs, but did not control for either undergraduate major or an individual’s precollege ambition.  Both Bowen and Bok (1998) and Fitzgerald (2000) introduced statistical controls for cognitive ability and major field of study, but not for differential tuition costs or ambition.  Most recently, Thomas (2000) controlled for cognitive ability and academic major and considered differential costs in the form of a debt/earnings ratio.  However, he included no ambition measures in his prediction of earnings.

            It is measures of individual ambition that are almost universally absent in investigations of the impact of college quality on earnings.[19]  This absence should probably come as no great surprise as measuring ambition in a way that predicts one’s future economic success is a nontrivial challenge.  Unfortunately, the inability to adequately specify an individual’s ambition in regression models does not prevent unmeasured or unobserved ambition from confounding the relationship between the selectivity of the college one attends and his or her subsequent earnings.  In short, elite, highly selective colleges may simply recruit and enroll students who would have a high earnings capacity, irrespective of where they went to college.

            This issue has been creatively addressed in an important study by Dale and Krueger (1999).  They hypothesized that, given broad public awareness of the link between attending an elite college and career success, the selectivity of the colleges to which a student applies may signal unaccounted for ambition and earnings capacity.  They tested this hypothesis for a combined sample of men and women by reanalyzing data from the 1995 follow-up of the 1976 cohort from the College and Beyond data set. (Previous analyses of this data by Bowen and Bok (1998) had yielded significant net positive effects of college selectivity on 1995 earnings.) A basic equation was developed which regressed the natural logarithm of 1995 earnings on predicted parental income, individual SAT score, sex, race, high school academic achievement, collegiate athletic participation, and college selectivity (average SAT score).  In this equation college selectivity had a significant, positive effect on earnings.  However, when measures of the average selectivity of the colleges to which one applied and the number of applications one made were added to this equation, the effect of college selectivity on earnings was reduced to a magnitude that was trivial and nonsignificant.  Furthermore, this finding appeared to be robust.  Almost exactly the same results were obtained when the same control for ambition was applied to a combined sample of men and women from the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972—a more nationally-representative sample than College and Beyond.  When a measure of precollege ambition was taken into account in either sample, students who attended more selective institutions did not earn more than their counterparts who were accepted and rejected by comparable schools but attended less selective institutions.

            Despite the volume of evidence concerning college selectivity and earnings reviewed above, we tend to agree with Dale and Krueger (1999, p. 29) that their “findings cast doubt on the view that school selectivity, as measured by the average SAT score of the freshmen who attend a college, is an important determinant of students’ subsequent incomes.”  Put another way, extremely bright and ambitious students (i.e., those with a high earnings capability) are more likely than other students to attend and graduate from highly selective colleges.  Whether such elite institutions contribute significantly more to those students’ earnings capabilities[20] than would less selective schools, however, is problematic.[21]

            In addition to the problem of unmeasured ambition, there is also the question of which students are really receiving significant economic returns from attending a selective institution.  There is at least some evidence to suggest that studies that fail to take into account a student’s educational path, practically a universal characteristic of the existing body of research, overestimate the effects of institutional selectivity for what may be the majority of students.  In an analysis of the 1986 follow-up of the nationally-representative High School and Beyond sample, Hilmer (2000) estimated the net returns to undergraduate college selectivity for three groups of male graduates: direct attendees (those who initially enrolled at the institution and remained there through graduation); university transfers (those who transferred to the institution from another four-year college); and community college transfers (those who transferred to the institution from a two-year community college).  With controls for such factors as race, high school and college grades, tested math and reading ability, college major, labor market experience, and having a postgraduate degree, but not ambition, the overall effects of college selectivity (average student SAT score) on earnings were trivial and nonsignificant for direct attendees.  Four-year and community college transfers derived significant economic returns from selectivity, but only if they transferred to, and graduated from, a four-year institution that had an average student SAT score of 1200-1400.  Since four-year and two-year college transfer students comprised only about a third of Hilmer’s nationally-representative High School and Beyond sample, his findings suggest that for the majority of male four-year college graduates (direct attendees) the selectivity of the college attended has little impact on their subsequent economic success.

Institutional Control (Private Versus Public)

Career Choice

            There is some limited evidence that institutional control may have an impact on the prestige of one's career choice.  In Astin's (1993b) analyses of the 1985-89 Cooperative Institutional Research Program data, he attempted to estimate the effects of different institutional characteristics on college seniors' choice of various careers.  With controls for initial career choice, other individual-level background traits, institutional characteristics, and measures of student academic and social involvement in college, attending a private university had a positive influence on seniors' choice of physician as a career.  Conversely, attending a private institution had a negative effect on choice of school teacher as a career.

Occupational Status

            We found little to suggest that attending a private (as compared to a public) postsecondary institution had anything more than a trivial and statistically nonsignificant influence on overall occupational status.  Analyzing the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972, Knox, Lindsay, and Kolb (1993) introduced statistical controls for race, sex, family background, tested academic ability, college grades, college major, educational attainment, institutional selectivity, and institutional size.  In the presence of such controls, attending a private (versus public) institutional as an undergraduate had only a small, chance influence on 1986 occupational status.  Consistent results are reported by Dey, Wimsatt, Rhee, and Waterson (1998) analyzing the 1974-75 and 1992-93 follow-ups of the Wisconsin Longitudinal Study of 1957 High School Seniors.
Earnings

            We uncovered a substantial body of studies that attempted to estimate the unique impact of attending a private (versus public) institution on earnings.  The specific data sets and studies that employ them are as follows:

1.     The National Longitudinal Study of the High School Class of 1972, 1986 follow-up (Arcidiacono, 1998; James & Alsalam, 1993; James et al., 1989; Knox et al., 1993; Sweetman, 1994a, 1994b).

2.     The National Center for Education Statistics Surveys of Recent College Graduates: 1985-86 graduates followed up in 1987 (Rumberger & Thomas, 1993); 1989-90 graduates followed up in 1991 (Tsapogas et al., 1994).

3.     The National Science Foundation New Entrants Survey of 1992 graduates followed up in 1993 (Tsapogas et al., 1994).

4.     The Baccalaureate and Beyond Study of 1992-1993 graduates followed up one year later (S Thomas, 1998).

5.     The National Longitudinal Survey of Youth, 1987-1989, and 1993 follow-ups (Daniel et al., 1996b; Monks, 2000).

6.     The High School and Beyond 1980 cohort followed up in 1986 (Fox, 1993).

7.     The Cooperative Institutional Research Program data, 1985 freshmen followed up in 1994 (Avalos, 1996).[22]

8.     A survey of identical and nonidentical female twins born in Minnesota and followed-up in 1993 at about age 45-46 (Behrman, Rosenzweig et al., 1996).

            With two exceptions (Avalos, 1996; Behrman, Rosenzweig et al., 1996), all of the studies cited above attempt to control, not only for individual student characteristics (e.g., race, sex, socioeconomic background, tested ability, and the like), but also for a measure of the academic selectivity of the institution attended.  Thus, the estimates they report are for attending and graduating from a private institution, irrespective of its level of selectivity.  About half of these studies find a small, but statistically significant, positive effect on earnings accruing to students who receive their bachelor's degree from a private college.  The other half find nonsignificant, but generally small positive effects.  Across all studies that provide requisite information, we estimate the average net earnings advantage associated with graduating from a private institution (irrespective of its level of selectivity) to be about 3%.  Due to the designs of the studies, however, this estimate doesn't discount the differential tuition costs between attending private versus public institutions.[23]  Were these differential tuition costs taken into account, the net earnings premium associated with attending a private institution would in all likelihood be considerably reduced, at least early in one's career.  Thomas (1998), for example, shows how attendance at a private (versus public) institution leads to a statistically significant and substantially higher debt to earnings ratio for recent college graduates.  The estimates also do not take into account individual precollege ambition or earnings capability, which might be differentially distributed across private and public institutions in much the same way it was differentially distributed across institutions differing in selectivity (Dale & Krueger, 1999).

            There is also limited evidence to suggest that different types of private colleges may differentially influence earnings, even when their level of selectivity is held constant.  For example, Sweetman (1994a; 1994b) found that only those private colleges that were not affiliated with a church had a significant positive influence on earnings.  James and Alsalam (1993) found that, at least for men, earnings were significantly enhanced by attending a private college only if the college was located in the northeastern United States.  These findings, however, are based on single samples and await replication.  Moreover, it is not clear if the private nonsectarian or private northeastern categories are really proxies for other institutional characteristics.

Career Eminence

            We uncovered one study (Wolf-Wendel, 1998) that estimated the effect of attending a private (versus public) institution on the career eminence or success of women.  Career eminence was operationally defined as inclusion in one of several national Who's Who compilations: Who's Who in America, Who's Who Among Black Americans, and Who's Who Among Hispanic Americans.  Institutions were the unit of analysis.  With statistical controls that varied for different analyses, but which typically included such institutional factors as selectivity, racial composition, and gender composition, obtaining a bachelor's degree from a private institution positively influenced career eminence for African-American and Latino women, but had no significant effect on the career eminence of white women.  Whether such findings for African-American and Latino women are attributable to socialization or recruitment effects, however, is somewhat difficult to discern from Wolf-Wendel's analyses.  It was not possible for her to control for such factors as the average entering aspirations, career ambition, or even family background characteristics of African-American and Latino women who enrolled in private (versus public) institutions.  Even with institutions as the unit of analysis, differences in these average entering student characteristics between private and public colleges may be a quite reasonable alternative explanation for her results.  As suggested by the work of Dale and Krueger (1999), controls for institutional selectivity may not provide an adequate proxy for student ambition.

Carnegie Classification

Earnings

            A modest body of research in the 1990s has attempted to estimate the effect of the Carnegie Classification of institutions on earnings.  The Carnegie Classification (Carnegie Foundation for the Advancement of Teaching, 1994) places four-year institutions in the following general categories: Research Universities (I and II, depending on annual number of doctorates awarded and external research funding); Doctoral Universities (I and II, depending on range of doctoral programs and number of doctorates awarded); Comprehensive Institutions (I and II, depending on range of master's programs and master's degrees awarded—these institutions have no doctoral programs); Liberal Arts Colleges (I and II, depending on selectivity); and Specialized Institutions (e.g., medical, engineering, business, and the like).  Unfortunately, these classifications are confounded by such factors as student body selectivity, private (versus public) control, institutional resources, and size.  Consequently, findings based on the Carnegie Classification might well represent proxies for other institutional characteristics.  For example, in analyses of the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972, Grubb (1992b; 1995b) sought to determine if the net economic returns to a bachelor's degree differed in magnitude for students graduating from different Carnegie Classification institutions.  Net of such factors as race, family income, high school grades, tested ability, and job experience, the returns (versus a high school degree) were relatively stable across the different institutional classifications, except for the categories having the most and least selective institutions.

            Across all the studies we uncovered that considered the economic returns to earning a bachelor's degree from different Carnegie-type institutions, only Specialized Institutions (e.g., those focusing on medical specialties, business, engineering, and the like) had a consistently positive net effect on earnings (Bellas, 1998; Monks, 2000; Tsapogas et al., 1994).  Moreover, this positive effect remained statistically significant, even when measures of institutional selectivity were taken into account (Monks, 2000; Tsapogas et al., 1994).  In all these studies, the comparison group was either Liberal Arts Colleges I and II grouped together (Bellas, 1998; Monks, 2000) or Liberal Arts Colleges II (Tsapogas et al., 1994).  The average earnings advantage accruing to graduates of Specialized Institutions (versus graduates of Liberal Arts Colleges) was about 19%.

            Evidence with respect to the relative economic returns linked to a bachelor's degree from other Carnegie-type institutions is markedly less consistent.  Monks' (2000) analyses of the 1993 follow-up of the 1979 cohort of the National Longitudinal Survey of Youth found that, even in the presence of controls for student background characteristics, work experience, and a measure of institutional selectivity, graduating from a Carnegie Research University, Doctoral University, or Comprehensive (Master's) University (versus a Liberal Arts College I or II) provided a positive and statistically significant advantage in earnings.  There was little support for this, however, in analyses of: the NCES Recent College Graduates Survey and the NSF New Entrants Survey (Tsapogas et al., 1994); and the 1991 follow-up of the High School and Beyond (HSB) 1980 cohort (Fitzgerald, 2000).  With similar controls in place, including measures of institutional selectivity, there were no significant earnings differences in the NSF New Entrants Survey between Liberal Arts II graduates, on the one hand, and graduates of Research, Doctoral, or Comprehensive Universities on the other.  In the NCES Recent College Graduates Survey, the net returns to graduates of Comprehensive I Universities were significantly higher than those of Liberal Arts Colleges II, but the net returns to graduates of Research I Universities were significantly lower.  The waters are muddied even further by Fitzgerald's (2000) findings that women receive an early career earnings boost by graduation from a Liberal Arts I College, while an earnings penalty accrues to men who graduate from a private Research I University.

Bellas' (1998) analyses of the 1993 Baccalaureate and Beyond study suggest that graduation from a Research or Doctoral University (versus a Liberal Arts College) has a modest, positive indirect effect through labor market experience and occupational classification.  Because she could not control for institutional selectivity, however, it is unclear how much this indirect effect might be confounded.[24]

            Although they do not place major focus on the Carnegie typology, three additional studies have estimated the economic premium associated with earning a bachelor's degree from an institution with a doctoral program.  In these studies, doctoral program institutions were compared with all others, but, like the studies reviewed above that employed the Carnegie classifications, the results were inconsistent.  Behrman, Rosenzweig, and Taubman (1996) found that receiving a bachelor's degree from a doctoral granting institution had a positive net influence on subsequent earnings, but both James, Alsalam, Conaty, and To (1989), and Hilmer (2000) report findings indicating that it did not.  Furthermore, Hilmer also reports an additional finding suggesting that graduating from a Research I University may have actually had a small negative impact on subsequent earnings.  In all three investigations, controls were introduced for institutional quality measures as well as individual-level student characteristics.  Only the James et al. and Hilmer studies also included a direct measure of institutional selectivity, however.

            Generally, then, when institutional selectivity is taken into account, it is questionable that either an institution's Carnegie classification or its doctoral/research orientation has a consistent, statistically significant link to an individual's subsequent earnings.  The one exception to this finding appears to be graduation from a Carnegie-Type Specialized institution; and this exception is likely because such institutions frequently focus on preparing individuals for occupational fields characterized by high economic returns (e.g., medical specialization, engineering, business).

Institutional Size

Occupational Status

            We uncovered little evidence in the research of the 1990s to suggest that institutional size has anything more than a small and statistically nonsignificant impact on occupational status.  Analyzing data from the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972, Knox, Lindsay, and Kolb (1993) introduced controls for such individual- and institutional-level variables as race, sex, socioeconomic status, college grades and major, institutional selectivity, and private (versus public) control.  In the presence of these controls, the size of the institution attended (operationally defined as student enrollment) had a small, positive, but statistically nonsignificant impact on 1986 occupational status.  Generally similar results are reported by Dey, Wimsatt, Rhee, and Waterson (1998) in their analyses of the 1974-75 and 1992-93 follow-ups of the Wisconsin Longitudinal Study of 1957 high school seniors.

Earnings

            A modest body of research has estimated the net effect of attending institutions of varying size (typically defined as student enrollment) on subsequent earnings.  With the exception of two studies, which report weak evidence for a statistically significant, negative effect (Behrman, Rosenzweig et al., 1996) or a very small nonsignificant, negative effect (Hilmer, 2000), the evidence consists of studies reporting that institutional size either has a significant, positive net influence on subsequent earnings (Avalos, 1996; Dowd, 1999; James & Alsalam, 1993; S Thomas, 1998), or a small and positive, but statistically nonsignificant net influence on earnings (James et al., 1989; Knox et al., 1993; Tsapogas et al., 1994).  Importantly, we believe, the studies that find a significant, positive impact of institutional size also introduce, in addition to controls for individual-level characteristics, controls for either institutional selectivity (Dowd, 1999), private control (Avalos, 1996), or both institutional selectivity and private control (James & Alsalam, 1993; S Thomas, 1998).  Indeed, Dowd found that graduates of large institutions (universities) that were substantially less selective actually had an earnings advantage of about 14-15% over graduates of the most elite liberal arts colleges, those with an average student body SAT score greater than 1300.  We conclude that the weight of evidence suggests that, other things being equal, institutional size confers a small, but statistically significant advantage in earnings.  (The nature of the results reported by different studies makes determination of the magnitude of the effect problematic.)  As suggested by Dowd, this positive effect of graduating from a large institution probably stems from economies of scale in providing diverse programs and major fields of study, as well as more extensive job networks due to larger alumni groups.  Furthermore, by means of the greater number of majors and preprofessional programs they offer, larger institutions typically have a wider range of links with occupational and economic groups in society.  Other factors being equal, this fact may afford larger institutions superior status-allocating capacity than smaller institutions (Pascarella & Terenzini, 1991).

Institutional Racial Composition

            In previous chapters in this book we have seen that African-American students attending historically Black colleges (HBCs) not only make content knowledge and intellectual gains that are equal to, if not greater than, their counterparts attending predominantly white institutions (PWIs), they also are more likely to complete a bachelor's degree.  Does graduation from HBCs confer any distinct advantages on African-Americans in their career?

Career Preparation and Occupational Aspirations

            A small body of evidence suggests that African-American students attending HBCs tend to believe they have made greater gains in preparation for a career and report a higher level of occupational aspirations than their counterparts attending PWIs.  For example, DeSousa and Kuh (1996) asked African-American students attending an HBC and a PWI to indicate the gains they felt they had made during college in vocational and career skills (e.g., acquiring knowledge and skills applicable to a job, gaining a range of information that might be relevant to a career).  HBC students reported making gains on the scale that were about .58 of a standard deviation larger than African-American students at PWIs.  Such a finding is consistent with that of Cole, Barber, Bolyard, and Linders (1999) who found that African-American students at HBCs scored significantly higher than their counterparts at PWIs on an index measuring the extent to which they focused on school as a means to an occupation versus getting a broad liberal arts education.  The designs of these studies, however, make it difficult to determine if these differences in self-reported gains in vocational and career skills, and focus on college's instrumental value in preparing one for a career, are attributable to the influence of attendance at an HBC or PWI.  They may merely reflect differences in the precollege career orientations of African-American students who choose to attend HBCs versus PWIs.

            The unique influence of attending an HBC on African-American students' occupational aspirations has been addressed by Allen (1992) and Wenglinsky (1996).  In Allen's study of African-American students at eight HBCs and eight PWIs, statistical controls were introduced for such factors as educational aspirations, class level, college and high school grades, sex, socioeconomic status, self concept, and the like.  In the presence of these controls, attending an HBC (versus a PWI) had a modest, but statistically significant positive, direct effect on a measure of the prestige and power dimensions of one's occupational plans.  HBC attendance also had a modest, positive indirect effect on occupational plans, transmitted through the positive effects of HBC institutions on students' social involvement during college.  Although no controls could be introduced for prior educational or occupational aspirations, consistent findings are reported by Wenglinsky (1996) in analyses of the more nationally-representative National Postsecondary Aid Study of 1990.  Similarly, Astin (1993b) reported that attending an HBC had a significant positive, net influence on seniors' choice of physician as a career.

Occupational Status, Earnings, and Career Eminence 

            Evidence concerning the impact of graduating from an HBC (versus PWI) on African-Americans' actual occupational and economic attainments is mixed.  For example, Ehrenberg and Rothstein (1994) analyzed data from the 1979 follow-up of the National Longitudinal Study of the High School Class of 1972 to determine if an HBC conferred an early occupational status or earnings advantage on African-Americans.  Controlling for gender, SAT scores, high school rank, educational attainment, parents' education and income, father's occupational status, and the unemployment level in one's state of residence, attending an HBC (versus PWI) had only statistically nonsignificant effects on 1979 occupational status and 1979 earnings.  Similar findings are reported by London (1998) for occupational status, though with a much smaller and focused sample.[25]  However, analyzing data from the same sample, but with a later follow-up (1986), Constantine (1994; 1995) reported findings suggesting that attendance at an HBC may indeed have a positive impact on African-Americans' subsequent wages.  She used two different samples to derive her estimates.  In the first, she attempted to reproduce the analytical model of Ehrenberg and Rothstein by restricting the sample to African-Americans in four-year institutions.  Controlling for such influences as high school achievement, tested ability, sex, family background, athletic participation, and region of the country, she found that attendance at an HBC (versus PWI) conferred a statistically significant advantage of about 11% in 1986 wages.  When attainment of a bachelor's degree was added to the equation the advantage dropped to about 8% and became nonsignificant, suggesting that part of the positive total impact of attending an HBC was indirect, transmitted through the positive effect of HBC attendance on African-American students' completion of a bachelor's degree.

            In Constantine's (1995) second sample, she included all African-American students, irrespective of whether they enrolled in a four-year institution.  She initially used background and other characteristics to predict, or model, three choices: no four-year college (i.e., high school or two-year college), attendance at a four-year HBC, or attendance at a four-year non-HBC (PWI).  Incorporating a term representing a correction for this choice or selection in her basic regression model yielded two important findings.  First, the negative selection term suggested that the unobservable characteristics that led a student to select an HBC (over a PWI) were probably the ones that would have caused lower wages.  Second, with this correction for selection taken into account, the estimated value-added in 1986 wages from attending an HBC (versus  PWI) was actually about 38%.

            Other research presents a less optimistic picture of the net impact of attending an HBC on African-Americans' career and economic success.  For example, Solnick (1990) examined the impact of attending an HBC on the job success of a sample of African-American college graduates employed by a large U.S. manufacturing firm.  Three job success outcomes were predicted: starting salaries, salary growth, and promotion.  With controls introduced for extensive personal and job characteristics, the resources of the colleges attended, and for possible attrition bias, African-American employees who were graduates of HBCs had a modest, but statistically significant 4% advantage in starting salary over their counterparts who graduated from PWIs.  Conversely, HBC graduates were significantly disadvantaged in both percentage of salary growth and the probability of being promoted within two years of being hired.

            In his analyses of the 1997 follow-up of 1993 bachelor's degree recipients in the Baccalaureate and Beyond study, Thomas (2000) estimated the net impact of attending an HBC on both annual earnings and the debt to earnings ratio.  Controlling for such factors as sex, race, family background, SAT scores, college grades and major, variables capturing labor-market experience, and both the selectivity and private/public control of the institution attended, earning a bachelor's degree from an HBC (versus all other institutions) had no significant impact on the 1997 debt to earnings ratio.  Graduates of HBCs did, however, have a statistically significant, 23% disadvantage in 1997 earnings.  Unfortunately, the sample used by Thomas includes individuals of all races, not just African-Americans.  Consequently, he is comparing graduates of HBCs, who presumably are nearly all African-Americans, with graduates of all other institutions, irrespective of race.  Thomas did, however, include dummy variables to represent African-American as well as Asian, Hispanic, and Caucasian (coded o) racial categories.  Thus, a reasonable interpretation of his findings is that for individuals in the same racial category (including African-Americans), the average net effect of graduating from an HBC is a 23% disadvantage in 1997 earnings.  Employing a similar sample and analytic design with the 1991 follow-up of the High School and Beyond 1980 cohort, Fitzgerald (2000) reported that graduation from an HBC had a statistically non-significant net effect on early career earnings of both men and women.  Obviously, the effect of HBC graduation on African-Americans' earnings would have been estimated more precisely with a sample limited to African-Americans.  Nevertheless, the findings of both Thomas and Fitzgerald are not irrelevant to the body of evidence.

            In addition to the body of evidence concerning earnings, Wolf-Wendel (1998) sought to estimate the net impact of graduating from an HBC on the career eminence of African-American women.  In the same study, she also estimated the impact on career eminence of Latino women attributable to earning a bachelor's degree from a primarily Hispanic serving institution (i.e., an institutional member of the Hispanic Association of Colleges and Universities--HACV).  Institutions were the unit of analysis and career eminence was defined as inclusion in one of two respective Who's Who compilations: Who's Who Among Black Americans and Who's Who Among Hispanic Americans.  Net of statistical controls for institutional control and gender composition, African-American women graduates of historically Black colleges were significantly over-represented in terms of career eminence.  Similarly, net of controls for institutional control, selectivity, and gender composition, graduating from a primarily Hispanic serving institution (i.e., a member of the HACV) had a significant, positive influence on the career eminence of Latino women.  As with Wolf-Wendel's (1998) findings concerning graduation from a private college, however, there is the very real possibility that these findings are also confounded by the inability to control for such factors as the average career aspirations, ambition, and family backgrounds of students entering HBCs and Hispanic serving institutions.

            Overall, it is difficult to form a firm conclusion about the impact of attending an historically Black institution on African-American's career and economic success.  Historically, Black colleges appear to enhance the career aspirations of African-American students; and there is some evidence that a bachelor's degree from an HBC is at least associated with one dimension of career eminence among African-American women.  However, the weight of evidence with respect to the influence of graduating from an HBC on African-American's occupational status, career mobility, and earnings is not totally convincing.  The study of economic returns that follows African-American students furthest in their careers (Constantine, 1995) also yields the most positive estimates of HBC attendance on earnings or wages.  Her findings are based on a single sample, however, and await replication.

            Although the vast majority of evidence on the effects of college racial composition on career and economic attainment focuses on historically Black colleges and African-Americans, we uncovered four additional studies that estimated the impact of an institution's student-body racial composition on the earnings of non-African-Americans as well as African-Americans.  Analyzing the four-year follow up of the 1993 Baccalaureate and Beyond study, Thomas (2000) introduced statistical controls for an extensive set of individual-level and institutional-level characteristics.  These included: race, sex, family income, college grades, college major, tested ability, section of the country, and labor market experience at the individual level, and selectivity, private (versus public) control, and attendance at an historically Black college at the institutional level.  In the presence of such controls, graduates of colleges with more diverse student bodies enjoyed a statistically significant earnings advantage relative to those from more racially homogeneous campuses.  On average, a 10% increase in non-white students on campus led to a 3% increase in 1997 earnings, net of other factors.  (Such results are quite similar to those found earlier for students majoring in business, education, and health-related fields by Rumberger & Thomas, 1993).  Thomas concludes that this finding may suggest that employers are recognizing and rewarding recent graduates' experiences with diverse populations, and these experiences are more likely to happen at institutions with diverse undergraduate student bodies (Gurin, 1999).

            Consistent, if not totally comparable findings are reported by Daniel, Black, and Smith (1996a; 1996b) in their analyses of the 1987-89 follow-up of the 1979 cohort of the National Longitudinal Survey of Youth.  In their investigations, they created four categories representing the percent of African-American students at each institution: less than 5%; 5-7%; 8-17%; and more than 17%.  Statistical controls were introduced for such individual-level influences as tested ability, home and family background, age, race, high school quality, college major, labor market experience, and industry of employment, and for a composite measure of college quality (e.g., selectivity, spending per student, faculty with Ph.D.s, and the like) at the institutional level.  Net of such controls, they found that men attending colleges with between 5 and 7% African-American students earned significantly more than those attending colleges with fewer than 5% African-American students.  Furthermore, men attending colleges with between 8 and 17% African-American students had significantly higher earnings than men at schools with fewer than 8% or more than 17% African-American students.  There was no significant difference in the magnitude of the effect of college racial diversity for African-American versus non-African-American men (Daniel et al., 1996b).  Thus, even with controls for college quality and background characteristics, attending a college with a moderate level of racial diversity (percent African-American) among its students significantly raised earnings for both African-American and non-African-American men, and did so about equally.

            The corresponding results for women were less clear (Daniel et al., 1996a).  Net of background characteristics and institutional quality, African-American women who attended colleges with between 5 and 7% African-American students earned significantly more than their counterparts who attended colleges with less than 5% African-American students.  The trends in the evidence also suggest they earned more than otherwise similar African-American women who attend colleges with more than 8% African-American students.  Percent African-American students at the institution attended had only small and statistically nonsignificant effects on the earnings of non-African-American women.

            Since it is unclear how many HBCs fell into the category of "more than 17% African-American students," it is difficult to draw a direct comparison between the findings of the Daniel, Black, and Smith (1996a; 1996b) studies and those that compare African-American students who attend HBCs with those who attend predominantly white institutions.  Nevertheless, the results reported by Daniel, Black, and Smith suggest that, even if an African-American student does not attend an HBC, a modest percentage of other African-Americans on campus may positively influence his or her subsequent earnings.  Furthermore, taken together, the findings of Thomas (2000) and Daniel, Black, and Smith (1996b) suggest the intriguing possibility that non-African-American students, and particularly men, may derive potential benefits from experiences on a racially diverse campus that translate into subsequent economic advantages.

Institutional Gender Composition

            As of the middle of the 1990s, there were fewer than 70 baccalaureate-granting women's colleges in the United States, and they granted slightly more than 2% of all bachelor's degrees awarded to women (College Entrance Examination Board, 1994; Ricci, 1994; Wolf-Wendel, 1998).  Thus, in terms of both numbers of institutions and numbers of graduates, single-sex women's colleges could be considered only minor players in the overall national postsecondary system.  In terms of the accomplishments and influence of their graduates, however, women's colleges are anything but minor players.  Graduates of women's institutions hold positions of leadership and eminence in such fields as government, business, the professions, and postsecondary education that are dramatically out of proportion to their small numbers (H. Astin & Leland, 1991; Forbes, 1998; Harwarth, Maline, & DeBra, 1997; Ledman, Miller, & Brown, 1995; Tidball, Smith, Tidball, & Wolf-Wendel, 1999; Touchton, Shavlik, & Davis, 1993).  The major question for social scientists interested in the impact of college is whether the marked accomplishments of women's college graduates are the result of some unique socialization process that goes on in women's colleges or merely reflect the recruitment of particularly talented and ambitious young women to those institutions.  A substantial body of research has addressed this issue.

Acquiring Career-Related Skills and Attitudes

            A number of scholars have been concerned with the extent to which attendance at a women's college influences career-related skills and attitudes.  Since women's institutions provide more opportunities for women to exercise leadership skills as well as interact with successful faculty role-models in a wide range of academic fields who are themselves women, it seems reasonable to hypothesize that attendance at a women's college would foster such outcomes as leadership skills, orientations toward success or accomplishment, and drive to achieve (e.g., Miller-Bernal, 1993; Romano, 1996).  With one exception (A. Astin, 1993b), however, the weight of evidence from the 1990s has failed to support this hypothesis.  For example, in a comprehensive and methodologically sophisticated investigation Smith, Wolf, and Morrison (1995) used the 1986-90 Cooperative Institutional Research Program data to compare women at 30 women's colleges with those who attended 173 private four-year coeducational institutions.  With statistical controls for individual-level precollege variables and SAT scores, institutional selectivity, and measures of academic and social involvement during college, attending a women's college had only a small and statistically nonsignificant direct effect on seniors' self-ratings of leadership ability and a scale measuring success goals and outcomes (e.g., have administrative responsibility, to become an authority in one's field, self-rating of drive to achieve, and the like).  Furthermore, the indirect effects of attending a women's college on the same two variables were trivial in magnitude.  Quite similar findings, using other iterations of the Cooperative Institutional Research Program data and essentially the same general analytical procedures as Smith, Wolf, and Morrison (1995) have been reported by Kim and Alvarez (1995) for the acquisition of job-related skills and preparation for graduate or professional school, by McKinney (1997) for self-assessed leadership ability, by Langdon (1997) for leadership ability and drive to achieve, and by Tullier (1990) for career salience and range of perceived career options.

Choosing Nontraditional Majors and Careers

            Since women's colleges (as compared to coeducational institutions) tend to provide a larger percentage of female faculty role models who function effectively in fields that are traditionally male-dominated and linked to high economic returns (e.g., economics, mathematics, natural sciences), it also seems reasonable to hypothesize that the environments of women's institutions may be particularly effective in counteracting sex-stereotypic perceptions in women's career aspirations and development (e.g., Riordan, 1994; Sebrechts, 1992; S. Solnick, 1995).  Here, too, the evidence is mixed.  On the one hand, we have the supportive findings of Sebrechts (1992) and Solnick (1995).  Summarizing a report from the Women's College Coalition, Sebrechts points out that women at women's colleges were three times as likely to earn a bachelor's degree in economics and one and one-half times as likely to earn bachelor's degrees in the life sciences, physical sciences, and mathematics than women at coeducational colleges.  Of course, such evidence could merely reflect precollege differences in the intended majors of women who enroll in women's and coeducational colleges.  More pertinent to the actual impact of women's colleges is Solnick's study of changes in women's majors from entrance to graduation at eight women's and seven coeducational colleges.  Depending on how broadly female-dominated majors (e.g., education, social work, social sciences, and the like) are defined, about 40% to 75% of women at the eight women's colleges who began in such majors shifted to neutral or male-dominated fields (e.g., mathematics, natural sciences, economics, and the like) during their college careers.  This shift compared to only about 25% of women at the seven coeducational schools.  Approximately 22% of women at both types of schools left male-dominated majors.

            Other evidence, however, is less supportive of the notion that women's colleges actually enhance the likelihood that women will choose nontraditional majors and careers (Dickson, 1990; Touchton, Davis, & Makosky, 1991; Tullier, 1990).  In contrast to Sebrechts (1992), Touchton, Davis, and Makosky (1991) summarized a report from the National Center for Education Statistics indicating that about the same percentage of women in women's colleges and coeducational colleges received bachelor's degrees in engineering, mathematics, and the physical sciences.  Similarly, both Dickson (1990) and Tullier (1990) found that when women's and coeducational colleges of about equal selectivity were compared, senior women at both types of institutions were largely indistinguishable in terms of their overall career choices and choice of a nontraditional (e.g., male-dominated) career.  The seeming contrast between these findings and those of Solnick (1995) may in part be explained by the fact that Solnick focused on major field of study, while Dickson and Tullier focused on actual career choice.

            The evidence is also mixed with respect to the net impact of women's colleges on women's actual entrance into nontraditional careers.  For example, in their analyses of the 1971-80 Cooperative Institutional Research Program data, Stoecker and Pascarella (1991) attempted to estimate the net effect of attendance at a women's versus a coeducational college on women's entrance into a nontraditional/male-dominated occupation.  Male-dominated occupations were operationally defined by the percentage of males within each occupational group, using data from the U.S. Bureau of Labor Statistics.  With statistical controls for such confounding influences as individual-level background traits and aspirations, institutional selectivity and size, college major, college grades, marital status, and educational attainment, attending a women's college had only small and nonsignificant direct and indirect effects on women's entry into male-dominated careers.

            Generally, if not totally, consistent results are reported in a 1994 reanalysis of data indicating that graduates of women's colleges are substantially overrepresented in a specific male-dominated occupation—that of physician (Crosby et al., 1994).  When different procedures and variables were employed to introduce previously absent controls for institutional selectivity, the effect of graduating from a women's college and becoming a physician either became nonsignificant or was substantially reduced in magnitude relative to other predictors.  Furthermore, even with controls for institutional selectivity, it is quite possible that the relationship between attending a women's (versus coeducational) college and becoming a physician is still confounded by differences between the two institutional types in the percentages of women who aspire to a career as a physician when they begin college.

            Other evidence is somewhat more supportive of the belief that women's institutions enhance the likelihood of women selecting a nontraditional career path.  Sharpe and Fuller (1995) used the Doctoral Records File maintained by the National Research Council to examine the physical science and engineering doctorate productivity of the baccalaureate institutions of the cohort of women who completed a bachelor’s degree between 1976 and 1986 and who earned a doctorate prior to 1992.  Taking institutional size and Carnegie Classification into account, the overall median physical science and engineering doctorate productivity was significantly higher for historically women’s colleges than for coeducational institutions.  Most of this effect was attributable to the over-representation of women’s college graduates in earning a doctorate in chemistry.  In both mathematics/computer science and physics/earth science, graduates of women’s colleges demonstrated no significant advantage over their counterparts with a bachelor’s degree from coeducational institutions.  As Sharpe and Fuller point out, however, it is risky to attribute any of their findings to the impact of institutional socialization.  It may simply be that women’s colleges tend to enroll a higher proportion of women who aspire to careers in the physical sciences when they begin postsecondary education than do coeducational institutions.  Without taking these precollege career dispositions into account, it is difficult to determine if the Sharpe and Fuller findings represent a socialization or a recruitment effect. 

Workforce Participation

            Analyses of two national data sets, the 1971-80 Cooperative Institutional Research Program data (Stoecker & Pascarella, 1991) and the 1986 follow-up of the National Longitudinal study of the High School Class of 1972 (Riordan, 1992; Rothstein, 1995), suggest that attendance at a women's college has little impact on the likelihood of women participating in the workforce.  With controls for extensive individual-level background characteristics, institutional selectivity and size, marital status, college major, and educational attainment, both Stoecker and Pascarella (1991) and Rothstein (1995) found that the percent of undergraduate women at an institution had only small and statistically nonsignificant effects on the likelihood of a woman being employed.  Riordan (1992) actually found that women's college graduates were less likely to be employed after obtaining a bachelor's degree than were women graduates of mixed gender colleges.  Moreover, in the Rothstein investigation women's workforce participation was also uninfluenced by the percent of female faculty at an institution.

Occupational Status and Earnings

            The weight of evidence with respect to the net impact of attending a women's college on occupational status and earnings is inconsistent and unconvincing.  For example, in analyzing data from the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972, Riordan (1992; 1994) found that, net of such influences as family socioeconomic status, tested ability, mental status, region of the country, and hours worked, attending a women's (versus coeducational) college had a small, but statistically significant positive effect on the status or prestige of a woman's occupation in 1986.  This, however, was not replicated in Riordan's (1993; 1995) analyses of the 1986 follow-up of the 1980 High School and Beyond cohort.  With statistical controls similar to those employed in his analyses of the NLS-72 data, attendance at a women's college reduced the likelihood of being married and increased one's occupational aspirations, but had only a small and statistically nonsignificant effect on the actual status of the job held by a woman in 1986.  Similarly, in their analyses of the 1971-80 Cooperative Institution Research Program data, Stoecker and Pascarella (1991) found that attendance at a women's college had only small and nonsignificant direct and indirect effects on a woman's 1980 job status when the influence of salient individual- and institutional-level variables was taken into account.

            The evidence pertaining to the impact of attendance at a women's college on subsequent earnings differs little from that pertaining to its effect on occupational prestige.  Riordan's (1994) analyses of the 1986 follow-up of the NLS-72 data present the strongest evidence of a positive impact.  Net of other influences, attending a women's college had a statistically significant, positive total effect on 1986 earnings.  Most of this positive effect of attendance at a women's college was indirect, being transmitted through a direct positive impact on 1986 occupational status which, in turn, positively influenced earnings.  There is a good chance, however, that this effect is biased upward, or inflated, because precollege (1972) occupational aspirations were not specified and, therefore, not controlled statistically, in Riordan's regression models.  Such precollege aspirations are important determinants of both subsequent occupational status and earnings.  For example, using essentially the same database as Riordan (1994), both Inoue and Ethington (1997) and Whitaker and Pascarella (1994) found that precollege (1972) occupational status aspirations had a significant positive effect on 1986 occupational status, even when precollege educational aspirations and subsequent educational attainment (among other variables) were taken into account.  Occupational status aspirations also had a significant positive effect on 1986 earnings, even in the presence of controls for educational aspirations, educational attainment, and occupational status (Whitaker & Pascarella, 1994). The importance of controlling for occupational aspirations is also underscored in Kingston and Smart's (1990) analyses of the 1971-80 Cooperative Institutional Research Program data.

            Additional evidence suggesting that attendance at a women's college may enhance subsequent earnings is provided in Dowd's (1999) ten-year follow-up study of graduates of 20 "highly prestigious" institutions.  With controls for undergraduate major, ethnicity, financial aid, geographic region, marital status, and dependent children, women graduates of three highly selective liberal arts colleges actually had significantly lower earnings than their counterparts at five less selective universities.  However, women graduates of four less selective liberal arts colleges, three of which were women's colleges, were not disadvantaged relative to the five comparison universities.  In terms of impact on subsequent earnings, Dowd reasoned that the distinctive culture of women's colleges may have compensated for the institutional size disadvantage of liberal arts colleges relative to universities.  (Recall our earlier review indicating a positive influence of institutional size on earnings.)  While such findings are intriguing, they are also based on an extremely small sample of institutions with quite limited generalizability.  Moreover, the results may be confounded by the inability to control for precollege levels of career aspirations and ambition.

            With the possible exception of Riordan (1994) and Dowd (1999), however, the weight of evidence is reasonably clear in suggesting that no statistically significant earnings benefit accrues to women who attend and/or graduate from a women's college versus a coeducational institution (Behrman, Rosenzweig et al., 1996; Daniel et al., 1996a; Riordan, 1993; Rothstein, 1995; Stoecker & Pascarella, 1991; Sweetman, 1994a, 1994b).  These studies tend to analyze data from nationally-representative samples such as the 8- and 9-year follow-ups of the 1979 cohort of the National Longitudinal Survey of Youth; the 1986 follow-up of the NLS-72 data; the 1986 follow-up of the 1980 cohort of the High School and Beyond data; and the 1971-80 Cooperative Institutional Research Program data.  An additional data set analyzed was a survey of identical and nonidentical female twins born in Minnesota and followed up in 1993 at about age 45-46.  The studies typically make attempts to introduce statistical controls for all or some of the following: individual-level background traits (e.g., tested ability, precollege educational or occupational aspirations, socioeconomic status); other institutional characteristics (e.g., selectivity, size); college experiences (e.g., academic major, grades); marital status and/or family obligations; educational attainment; and labor-market experience (e.g., work experience, sector of employment).  With such controls in place, attending a women's college, or the percent of undergraduate women at the college attended, had no statistically significant positive direct or indirect influence on a woman's subsequent annual earnings or hourly wages.  One of the studies we reviewed (Daniel et al., 1996a) suggests the possibility that attendance at a women's college actually confers a net earnings disadvantage on women.  However, considering the evidence presented across the entire body of studies we uncovered, it is difficult not to conclude that women's colleges confer neither an earnings benefit nor an earnings penalty on women.[26]

            While they are each based on findings from a single study, and await replication, two other findings are worthy of mention.  First, although Daniel, Black, and Smith (1996a) report that attending a women's college had a net negative effect on a woman's earnings, this was not the case for the economic productivity of a woman's spouse.  Net of other influences, the percent of women at the institution attended had a significant, positive effect on the earnings of a woman's spouse. [Similar findings are reported by Riordan (1992) for husband's occupational status, suggesting that the results of Daniel, Black, and Smith are not merely fortuitous or artifactual.]  Second, Sweetman's (1994a; 1994b) analysis of the 1986 follow-up of the NLS-72 data found that, net of other factors, attending an all-male college (versus coeducational institution) conferred a 23% earnings advantage on white men.

Career Eminence

            We uncovered only one study (Wolf-Wendel, 1998) that estimated the net impact of attending a women's college on a woman's career eminence.  Institutions (versus individuals) were the unit of analysis, and career eminence or success for women was defined as earning a bachelor's degree after 1965 and being listed in one of three Who's Who compilations—Who's Who in America, Who's Who Among Black Americans, and Who's Who Among Hispanic Americans.  Separate analyses were conducted for three groups of women—White/European Americans, African-Americans, and Latinos.  Wolf-Wendel essentially was trying to predict the proportion of women baccalaureate graduates from four-year institutions included in these Who's Who compilations.  Net of such institutional characteristics as selectivity, size, and private control, graduates of women's colleges were dramatically overrepresented in the three Who's Who compilations.  Clearly, Wolf-Wendel's study is limited by the fact that she could not control for differences in the average levels of precollege aspiration and ambition among women attending women's and coeducational colleges.  To some extent, her findings could simply reflect differential recruitment of ambitious women to women's and coeducational colleges, rather than any unique career socialization that might occur at the former.  On the other hand, it is extremely difficult to argue with the magnitude of Wolf-Wendel's estimates of the impact of women's colleges.  The standardized regression coefficients for institutional gender in her analyses vary between .27 and .59.  Part of these large effects, of course, could be attributable to the use of institutions rather than individuals as the unit of analysis.  Nevertheless, even if institutional-level controls could have been introduced for entering student ambition and aspirations, it is still questionable if Wolf-Wendel's estimates of the effects of women's colleges on women's career eminence would be reduced to statistical nonsignificance.

Two-Year Versus Four-Year Colleges

            In the Net-Effects of College section of this chapter, we reviewed the rather unsurprising evidence that, on average, completion of a bachelor's degree returned significantly higher levels of occupational status and earnings than did an associate's degree.  In this section, we review studies that focus on whether starting postsecondary education at a two- versus four-year college has important implications for one's career.
Persistence in Mathematics, Science, and Engineering

            We uncovered two longitudinal studies (Grandy, 1998; Hilton, Hsia, Cheng, & Miller, 1995), analyzing the same data, that estimated the net impact of starting postsecondary education at a two- versus four-year institution.  The sample, developed from data collected by the Educational Testing Service, consisted of 3,840 high-ability minority students (American Indian, African-American, Mexican-American, and Puerto Rican) who in 1985 scored at least 550 on the SAT-mathematics test and who indicated that they planned to major in mathematics, science, or engineering (MSE) in college.  (MSE majors were operationally defined as: agriculture, architecture, bio-sciences, computer sciences, engineering, medical and dental professions, mathematics, and physical sciences).  The sample was followed up in 1990 to determine whether or not they persisted in mathematics, science, or engineering.  MSE persistence in 1990 in the Hilton et al. study was defined as: 1) receiving a bachelor's degree in MSE and being engaged in full-time MSE work, 2) being enrolled full time in an MSE graduate school (regardless of what their undergraduate major may have been), or 3) having a bachelor's degree in an MSE field and being enrolled part time in an MSE graduate program.  In the Grandy study, MSE persistence in 1990 was defined as working or studying, full- or part-time (graduate or undergraduate), in an MSE field.  With statistical controls for such factors as tested verbal and mathematics ability, educational aspirations, family socioeconomic status, high school math and science experiences, gender, college grades, commitment to science during college, and the like, starting postsecondary education at a two-year (versus four-year) college had a statistically significant, negative direct effect on 1990 MSE persistence in both studies and a statistically significant, negative (though reduced) total effect on MSE persistence in the Grandy study.

Job Performance

            We uncovered only one study (Banta & Associates, 1993) that compared the job performance of two- and four-year college graduates.  A consortium of institutions in Tennessee surveyed their alumni about their employment status and satisfaction.  Employers were also identified and surveyed regarding salient employee traits.  Although employers of four-year college graduates perceived that their employees had more of the characteristics they expected of a college graduate, no statistically significant differences in employer perceptions of two- and four-year college graduates were found on global ratings of job performance, quality of general education preparation, or quality of education in the area of specialization.  Since it is unclear that graduates of two- and four-year colleges were holding the same level jobs, however, it does not necessarily follow from this finding that employers regard two- and four-year college graduates as equally competent employees.

Occupational Status and Earnings

            A small body of research has estimated the net impact on subsequent occupational status and earnings of starting postsecondary education at a two-year community college versus a four-year institution.  In terms of statistical significance, the findings with respect to occupational status are mixed, but they are quite similar in terms of the magnitude of the effect.  Monk-Turner (1990) analyzed data from the National Longitudinal Survey of Labor Market Experiences on men and women who were full-time workers in 1978 and were between 24 and 34 years old.  The dependent measure was occupational status or prestige as measured by the Socio-Economic Index or SEI.  In the presence of controls for age, a measure of mental ability, socioeconomic background, work experience, educational attainment, race, sex, marital status, and region of the country, starting postsecondary education at a two-year college (versus four-year college) resulted in a statistically significant disadvantage in 1978 occupational status of 2.83 points.  On the other hand, using the same dependent variable as Monk-Turner, analyses of the 1986 follow up of the National Longitudinal Study of the High School Class of 1972 (NLS-72) data by Whitaker and Pascarella (1994) led to a slightly different conclusion.  Controlling for sex, race, family socioeconomic status, age, secondary school grades, and extracurricular involvement, 1972 self-esteem, 1972 educational aspirations, 1972 occupational aspirations, college grades, and educational attainment, starting at a two-year college was associated with a statistically nonsignificant disadvantage in occupational status of 1.13 points.  Thus, across both studies the direct disadvantage in occupational status attributable to community college attendance was quite modest, about 2 SEI points or .10 of a standard deviation (4 percentile points).[27]

            In addition to estimating the impact of entering postsecondary education at a two- versus four-year college on subsequent occupational status, Whitaker and Pascarella (1994) also estimated the corresponding impact on 1986 earnings.  To the controls used in their prediction of 1986 occupational status (e.g., race, gender, socioeconomic status, 1972 aspirations, 1986 educational attainment, and the like), they added 1986 occupational status and hours worked per week.  Net of these influences, starting postsecondary education at a community college, versus a four-year institution, had only small and statistically nonsignificant effect on 1986 earnings.  Moreover, even without controls for 1986 educational attainment, occupational status, and hours worked per week, the effect of attending a community college on 1986 earnings was still small and nonsignificant.  Analyzing the same data, Adelman (1992, February; 1994) reports similar findings with respect to the impact of attendance at a community college on both earnings and home ownership.

Two-Year College Attendance and Transfer Institution "Selectivity"

            Although it does not speak directly to the impact of initially attending a community college versus a four-year institution on career success or earnings, it is worth briefly reviewing the findings of an additional study that speaks to the academic selectivity of the four-year institutions which community college transfers attend.  As we observed in reviewing the evidence, there are certainly legitimate questions about the magnitude of the impact of four-year college selectivity on an individual's economic success.  Yet, of all the institutional characteristics considered, the selectivity of the undergraduate student body at a college probably had the most consistent positive influence on a graduate's economic attainment.  From this perspective, Hilmer's (1997) creative study is of some relevance.  Analyzing combined data from the sophomores and seniors in the 1980 High School and Beyond study, Hilmer sought to determine if students who transferred from community colleges to four-year institutions end up at four-year institutions that are higher or lower in selectivity (operationally defined as the average combined SAT Verbal and Mathematics score for the institution's 1984 freshman class).  Educational path equations were developed that estimated the selectivity of the four-year institution attended based on a student's sex, ethnicity, high school program (college preparation or other), high school geographic region, family income, high school extracurricular activities, tested ability, and number of institutions per thousand students and fees charged by institutions in the student's home state.  Taking these influences into account, he predicted that a student who initially attended a community college transferred to a four-year institution that had an average student-body selectivity 32 SAT points higher that the four-year institution he or she would have attended right out of high school.  In short, the results suggested that students are able to attend more selective four-year institutions if they first attend community colleges.  Additional findings suggested that the predicted institutional selectivity benefit is largest for community college students who came from poor families, were of low tested ability, or performed poorly in high school.  Students whose family wealth, test scores, or high school grades were more than one standard deviation below the mean transferred to four-year institutions that were up to 75 SAT points higher than they would have attended right out of high school.  Conversely, high ability, high income, and high performing students lost little or nothing in terms of institutional selectivity if they decided to transfer.

Impact of Peers

Major and Career Choice

            In our 1991 synthesis, we concluded that the most consistent college environmental impact on career choice was that of "progressive conformity."  Progressive conformity posits that, other things being equal, a student's major field of study and career choice will be influenced in the direction of the dominant peer groups at an institution.  Our present synthesis found considerable evidence in support of the progressive conformity hypothesis.  Probably the most extensive and methodologically rigorous research done in this area has been carried out by scholars affiliated with the Higher Education Research Institute at UCLA (A. Astin, 1993b; A. Astin & Astin, 1993; Sax, 1994, 1996).  These scholars have analyzed various longitudinal iterations of the Cooperative Institutional Research Program data and have typically introduced statistical controls for an extensive set of individual-level characteristics, such as tested ability, race, sex, expected major, high school experiences, family background, precollege career plans, and measures of the academic and social experience of college.  In the presence of such controls, there is clear evidence that both major choice and career choice are influenced by the distribution of student majors at the institution attended.

            For example, in their national study of the factors that influence students' interest in studying science, mathematics, and engineering (SME) and pursuing careers in those areas, Astin and Astin (1993) found that, even after entering student characteristics (including initial choice of major) and other environmental variables had been controlled, a student's final major in four areas—biological science, physical science, engineering, and social science—was significantly and positively influenced by the percent of undergraduate peers at his or her institution majoring in those respective areas.  Similar findings have been reported by Astin (1993b) for senior-year career choice and by Sax (1996) for enrollment in science, mathematics, and engineering graduate programs.  Net of other factors, including precollege career choice, Astin (1993b) found that 1) a senior's career choice in business was positively influenced by the percentage of business majors at the institution, 2) a senior's career choice in engineering was positively influenced by the percent of engineering majors, 3) a senior’s career choice as a lawyer was positively influenced by the percent of social science majors, 4) a senior career choice to become a research scientist was positively influenced by the percent of natural science majors, and 5) a career choice as a school teacher was positively influenced by the percent of education majors.  Using a similar analytic design in her study of post-college commitment to science careers, Sax (1996) reported that enrollment in science, mathematics, and engineering graduate programs was significantly enhanced by attendance at an undergraduate institution where one's peers had a strong science orientation.  Peer science orientation was operationally defined as the percent of students at an institution initially choosing a career as a scientific researcher or college teacher plus the average importance to peers of making a theoretical contribution to science as a life goal.[28], [29]


Within-College Effects

Conclusions from How College Affects Students

            The evidence is clear that certain major fields of study (e.g., business, engineering, technical or professional) tend to have a closer fit with the skills required in one's first job than do others (arts, humanities, and social sciences).  It is not clear, however, that the job fit of one's major is a key determinant of job satisfaction.

            We found little evidence across studies that academic major, as typically categorized (humanities, social sciences, natural sciences, and so on), has more than a small and inconsistent pattern of effects on job status.  This may be due to the fact that traditional categorizations of major have only a marginal theoretical and functional fit with the structure of occupational status.  There is some modest evidence to suggest that when academic majors are placed on a continuum in terms of how they are "targeted" toward occupations that stress prestige, supervisory authority, or income, they demonstrate a stronger impact on job status.

            Although sparse, the evidence that college major independently influences the likelihood that women will enter sex-atypical careers is convincing.  Net of other factors, a sex-atypical major (one that attracts a high percentage of men, such as business or mathematics) enhances the likelihood of a woman entering a sex-atypical career.  Thus, academic major in college may be an important determinant of gender equality in the work force.  A student's major field of study, however, may have little to do with his or her job performance or long-term career mobility, although in the private sector this may depend upon the employing company.  Not a great deal of evidence pertains to these issues, but the evidence that does exist suggests that over the long run, in business at least, liberal arts majors do as well as (though not better than) those with a business or engineering degree.

            According to clear and consistent evidence, major field of study has a significant impact on early career earnings that cannot be accounted for by differences in the characteristics of students selecting different majors.  The majors that enhance earnings tend to be characterized by a relatively well-defined body of knowledge and skills, an emphasis on scientific or quantitative methods of inquiry, and often an applied orientation.  Examples include such majors as engineering, business, several of the physical sciences, and preprofessional majors oriented toward medicine and dentistry.  These majors tend to have close links to occupations with relatively high average earnings.  Differences in the academic field of study chosen during college tend to explain part but not all of the lower earnings of women and racial minorities.

            Nearly all of the studies on the influence of academic major on earnings focus on earnings during the early career.  The evidence is less convincing that the same majors are linked with higher earnings in the later stages of one's career.

            The evidence is consistent in suggesting that academic achievement during college has a small but statistically significant positive impact on early occupational status.  Part of this effect may be indirect, occurring because grades enhance educational attainment, a key determinant of job status.  Though less extensive, there is similar evidence to suggest that college grades also enhance the likelihood of women entering sex-atypical careers.  A substantial part of this influence may also be indirect, mediated through educational attainment.

            It was estimated that without other factors being controlled, college grades account for no more than 2 or 3 percent of the variance in various noneconomic indexes of job performance and career mobility.  Evidence from the most vigorously conducted study suggests that at least part of the link with career mobility may be causal.  This does not appear to be the case for the link between college grades and job satisfaction, however, which we interpret at spurious.

            The weight of evidence from a large body of research indicates that academic achievement during college has a positive direct impact on early career earnings that is independent of student background characteristics, the selectivity of the institution attended, and major field of study.  Evidence with respect to a longer-term effect is less extensive and inconsistent.  Any direct causal impact of grades on early career earnings appears small, probably explaining no more than 1 percent of the differences in individual earnings.  This might be increased by as much as one-third if the indirect effect of grades on earnings, through educational attainment, were also taken into account.

            We found no consistent evidence to suggest that extracurricular involvement during college has more than a trivial, net influence on the status of one's occupational choice, the actual occupational status of one's job, and one's earnings.  There is some limited support, however, for the contention that social leadership involvement during college enhances the likelihood of women entering sex-atypical careers.  With the exception of individuals in technical fields such as engineering, college graduates are consistent in indicating that extracurricular involvement, particularly in leadership roles, has a substantial impact on the development of interpersonal and leadership skills important to job success.  It also appears to be positively linked with managerial potential.  Objective assessments, however, indicate only a trivial link between career mobility and both the extent of extracurricular involvement and involvement in leadership positions during college.

            There is also little consistent evidence to suggest that intercollegiate athletic participation has anything but a trivial and statistically nonsignificant impact on occupational status or earnings, though it may enhance the social mobility of individuals from low socioeconomic backgrounds.

            The existing evidence suggests that, net of other influences, working during college, particularly in a job related to one's major or initial career aspirations, enhances the level of professional responsibility attained early in one's career, the likelihood of women choosing a sex-atypical career during college, and women's plans for entering the work force subsequent to college.  However, the evidence with respect to the influence of work during college on subsequent earnings is inconsistent.  This may be at least partially due to the fact that the studies reviewed do not typically consider the degree to which work during college is related to an individual's postcollege employment.

            The magnitude of faculty impact on student career choice appears to vary with amount of informal contact or interaction.  Net of other factors, including initial career choice, frequency of informal contact with faculty appears to enhance women's interest in a career as well as their choice of a sex-atypical career.  Similarly, it also has a net positive influence on orientation toward a scientific or scholarly career and, for some students, the status of one's career choice.  Although there are some problems in the designs of the extant studies, the weight of evidence also suggests that female faculty may be somewhat more influential career role models for women students than are male faculty.

Evidence from the 1990s

            A substantial body of literature in the decade of the 1990s has addressed within-college effects on dimensions of career and economic attainment.  We have organized our synthesis of that literature within the following general categories: interventions to enhance career development, academic major, academic achievement, extracurricular/social involvement, work during college, and academic involvement.
Interventions to Enhance Career Development

            There is voluminous literature on the career development of college students, most of which is either tangential to or beyond the scope of this synthesis.  However, there is a modest body of literature that estimates the impact of interventions designed to enhance students' career development during college.  This literature is largely experimental or quasi-experimental in design, but the nature of the interventions is not always clearly or comprehensively described.  Similarly, the dependent measures employed appear to assess a considerable range of career development dimensions.  Consequently, it is somewhat difficult to synthesize the findings.

            Fortunately, we were able to uncover at least one meta-analysis of career development courses.  Hardesty (1991) conducted a meta-analysis of 12 studies that attempted to evaluate the effectiveness of undergraduate career development courses that were offered for academic credit.  Thus, these courses were typically at least a semester in length and combined a range of didactic, experiential, and counseling activities.  In each of the studies, the outcome measure assessed either career maturity, career decidedness, or both.  Essentially, level of career maturity represents an individual's ability to make a realistic career decision while career decidedness appears to represent a person's level of certainty in their career choices.  Across all relevant studies, students enrolled in the career development courses demonstrated improvement in career maturity that was between .43 and .44 of a standard deviation (17 percentile points) greater than similar studies not exposed to the courses.  For career decidedness, the improvement advantage for students in the career development courses averaged between .34 and .36 of a standard deviation (13 to 14 percentile points).

            More recent literature would not appear to contradict Hardesty's (1991) major conclusion that career development courses or related interventions can significantly enhance dimensions of students' career development and maturity (e.g., Eveland, Conyne, & Blakney, 1998; Mau, Calvert, & Gregory, 1997; Niles & Garis, 1990; Sullivan & Mahalik, 2000; Wei-Cheng, 1999; Zagora & Cramer, 1994).  In some of these investigations, the intervention conditions consist of different computer-assisted or standardized career development programs such as: Career Decision-Making, a computer-assisted instructional program that teaches theory-based strategies for choosing a career; the Self-Directed Search, a vocational assessment and intervention designed to increase self-knowledge and the number of vocational options considered; DISCOVER, a computer-assisted program providing information about the fit between personality and potential careers; and the System of Interactive Guidance and Information Plus, an interactive system designed to help students clarify educational and career plans (Mau et al., 1997; Wei-Cheng, 1999).  In other studies, the intervention is a combination of computer-assisted interventions and career counseling or instruction (e.g., Eveland et al., 1998; Niles & Garis, 1990).  Still other studies employ different group instructional/counseling workshop formats as the experimental treatment (e.g., Sullivan & Mahalik, 2000; Zagora & Cramer, 1994).  Generally, the results of these investigations suggest that students in the various career development interventions (or combinations of interventions) show significantly greater growth than students not exposed to the interventions on a range of important career development dimensions.  These dimensions include: career decision-making efficacy, level of vocational exploration and commitment, vocational identity, vocational construct integration, and number of occupations considered.  Not all of the studies report requisite statistical information for computing effect sizes.  Moreover, given the differences across studies in the interventions employed and career-development outcomes assessed, it is not clear what an average effect size would have represented.  Consequently, we did not estimate an effect size for this more recent body of evidence.

            It is also worth briefly discussing a creative study by Luzzo and colleagues (Luzzo, 1995; Luzzo, Funk, & Strang, 1996) which suggests that attributional retraining as developed by Perry and his colleagues (Perry, Menec, & Struthers, 1996; Perry & Penner, 1990; Perry & Struthers, 1994) can increase career decision-making self-efficacy among certain kinds of students.  Recall from Chapter 3 that attributional retraining is an intervention strategy designed to enhance motivation and achievement striving by changing how students think about the causes underlying their success or failure.  There is evidence that attributional retraining is most effective in improving the learning of students who tend toward an external locus of attribution for success (e.g., attribute it to luck).  In the experiment by Luzzo et al., students were assigned either to a control condition or to an attributional retraining intervention.  In the latter, students watched a brief attributional retraining videotape in which two college graduates persuaded students to attribute career-related difficulties to a lack of effort and to attribute successful career development to adequate effort and persistence.  The dependent variable was career decision-making self-efficacy, or the extent to which a person feels confident in accomplishing tasks necessary to make good career decisions.  Consistent with the effects of attributional retraining on learning, the career decision-making self-efficacy of students with an external locus of control increased significantly after receiving the intervention while the career decision-making self-efficacy of students with an internal locus of control did not.

Academic Major

            With the possible exception of a student's academic achievement during college, one's academic major or major field of study is the most studied of all within-college effects on career and economic attainments.  As will become clear in our synthesis, undergraduate academic major can play a major, if not always totally consistent, role in one's career.  Moreover, academic major has important implications for gender equality/inequality in earnings.

Job-Related Skills

            A small body of research has addressed the issue of whether different academic majors have a differential impact on the development of job-related skills (Smart, 1997; Smart, Feldman, & Ethington, 1999).  Smart and his colleagues analyzed multi-institutional data from the 1986-1990 Cooperative Institutional Research Program sample and sought to determine if Holland's (1985) model of vocational choice was related to differential patterns of student growth.  Holland's model postulates that students with different personality characteristics, career goals, and aspirations tend to select academic subenvironments (or major fields of study) consistent with these precollege characteristics and goals.  In turn, these academic subenvironments tend to accentuate initial personality characteristics and career goals (e.g., Huang & Healy, 1997; Osipow & Fitzgerald, 1995; Smart & Feldman, 1998).  Thus, from the perspective of Holland's theory, one would hypothesize that students in "Enterprising" subenvironments (i.e., majors in such fields as business, marketing, pre-law, industrial engineering, and public administration) would demonstrate greater gains in practical, job-related skills than students in other clusters of academic majors (e.g., "Investigative": economics, natural sciences, experimental psychology, engineering; "Artistic": drama, English, philosophy, art, music; "Social": history, political science, sociology, nursing, education).  With statistical controls for students' precollege self-reported levels of competence, this is essentially what Smart and colleagues found.  Seniors in majors within the "Enterprising" subenvironment reported greater net levels of job-related skills and leadership growth (e.g., job-related skills, leadership skills, competitiveness, ability to work cooperatively, interpersonal skills) than did seniors in other clusters of majors (i.e., "Investigative," "Artistic," and "Social").

Workforce Participation

            There is little in the evidence from the 1990s to suggest that academic major has a statistically significant net impact on women's workforce participation (Bowen & Bok, 1998; Stoecker & Pascarella, 1991).  Bowen and Bok analyzed the College and Beyond data which followed up students about 19-20 years after entering college in 1976.  Their operational definition of major consisted of the categories of social science, natural science, engineering, humanities, and other.  With controls for such factors as race, SAT score, high school achievement, socioeconomic status, the selectivity of the institution attended, college academic achievement, educational attainment, marital status, and having children, undergraduate major had only a trivial impact on a woman's decision to work.  Similar, if not totally comparable, results are reported by Stoecker and Pascarella in their analyses of the 1980 Cooperative Institutional Research Program follow-up of students who began college in 1971.  Their operational definition of academic major was the percent of men in each respective field of study nationally in 1976.  Net of such influences as socioeconomic status, secondary school academic and social accomplishment, precollege educational and occupational aspirations; measures of institutional selectivity, size, and gender distribution; college academic achievement, marital status, and educational attainment, being in a sex-atypical or male-dominated major had a statistically nonsignificant impact on full-time labor force participation.

            One possible reason why academic major had no significant impact on workforce participation in the Bowen and Bok (1998) and the Stoecker and Pascarella (1991) studies is that there was simply too great a period of time between the undergraduate experience and the follow up—about 15 years for Bowen and Bok and about 5 years for Stoecker and Pascarella.  It may simply be that the impact of academic major on workforce participation is manifest early in an individual's career (e.g., Steinberg, 1994).  Some indirect evidence for this possibility is suggested in studies by Bellas (1998) and Sagen, Dallam, and Laverty (1997).  Bellas analyzed data from the first year follow-up of the 1993 Baccalaureate and Beyond Longitudinal Study and sought to predict the number of job interviews and job offers received.  Statistical controls were introduced for such factors as age, sex, race, marital status, institutional type, college grades, educational attainment, and work experience.  In the presence of these controls, business/management and engineering majors received, on average, the most job interviews while mathematics/computer/physical science majors and social science majors received the second most.  Net of the same controls, plus number of job interviews received, majors in business/management and health professions received the most job offers while engineers actually received the least.  The rather counterintuitive nature of the latter finding, given the demand for engineers, may be explained by the fact that engineering majors received lucrative offers early in their job search.  Thus, they may conclude their job search before some job interviews culminate in job offers (Bellas, 1998).

            Sagen, Dallam, and Laverty (1997) take a somewhat different approach and attempt to predict success in securing employment appropriate to the bachelor's degree within two months following graduation from a large, Midwestern, research university.  Employment appropriate to a bachelor's degree was determined by educational level in the Dictionary of Occupational Titles.  Major was operationally defined in terms of four categories: specialized-hard (e.g., engineering, computer science), specialized-soft (e.g., nursing, social work, education), broad professional (e.g., journalism and business), and general liberal arts (e.g., English, humanities).  Statistical controls were introduced for such potentially confounding influences as: ACT composite score, college grades, gender, work experience, coursework, volunteer activities, having a mentor as an undergraduate, and participation in student organizations.  In the presence of such controls, one's academic major significantly influenced the likelihood of appropriate employment.  Compared to general liberal arts majors, students in either the specialized-hard or specialized-soft majors were significantly more likely to have secured employment appropriate to a bachelor's degree.  Students in broad professional majors were also advantaged in securing appropriate employment over general liberal arts majors, but the advantage was not statistically significant.

            The findings of the Bellas (1998) and Sagen, Dallam, and Laverty (1997) studies are not totally consistent, in part perhaps because they address somewhat different outcomes.  The evidence they provide, however, does suggest that one's major can have a significant net impact on getting a job and securing employment at a level appropriate to a bachelor's degree early in one's career.  The clearest advantage in these areas would appear to accrue to students majoring in fields that have the most direct functional linkages with specific jobs or occupational sectors (e.g., computer science, engineering, social work, nursing, and perhaps some specific business fields such as accounting).

Occupational Status

            Unlike the conclusion from our 1991 synthesis, we found evidence in the research of the 1990s to suggest that academic field of study did have a significant, net impact on an individual's subsequent occupational status (Dey et al., 1998; Knox et al., 1993; Stoecker & Pascarella, 1991).  In all of these studies, occupational status was operationally defined with the Socio-Economic Index (SEI).  Dey et al. analyzed data from the 1974-75 and the 1992-93 follow-ups of the 1957 Wisconsin Longitudinal Study.  With statistical controls for such influences as gender, socioeconomic status, academic ability, high school rank, the characteristics of the undergraduate institution attended, and educational degree attainment, students with undergraduate majors in engineering, the health-related fields (e.g., pre-medical, pre-dental, pharmacy, nursing), and mathematics/science tended to be in jobs in the 1974-75 follow-up that had the highest occupational status.  Even when 1974-75 occupational status was added to the prediction equation, undergraduate majors in engineering and health-related fields continued to hold jobs with the highest occupational status in 1992-93.

            With the exception of such fields as nursing, the findings of Dey et al. (1998) suggest that students majoring in fields of study that have been traditionally dominated by men [e.g., engineering, mathematics, physical science, and technical pre-professional fields such as pharmacy, pre-medicine, and pre-dentistry; see, for example, Jacobs, (1995; 1996a); Nelson & Dixon, (1997); Turner & Bowen, (1999)] tend to be overrepresented in high status occupations.  Some generally corroborating evidence for this finding is reported by Stoecker and Pascarella (1991) in their analyses of the 1971-80 Cooperative Institutional Research Program data and by Knox, Lindsay, and Kolb (1993) who analyzed the 1986 follow-up of the National Longitudinal Study of the High School Class of 1972.  Recall that Stoecker and Pascarella operationally defined major as a continuous variable reflecting the percent of men in each respective field of study nationally in 1976.  Net of controls for such factors as precollege demographic characteristics, educational and occupational aspirations, characteristics of the undergraduate institution attended, college grades, and educational attainment, the percent of men in a woman's major field of study had a modest, but statistically significant positive effect on the occupational status of the job she held in 1980.  Similarly, Knox, Lindsay, and Kolb (1993) found that majoring in education or the liberal arts and sciences had a significant negative effect on the occupational status of one's job relative to majoring in engineering, technical, and professional fields.  This negative effect persisted even in the presence of statistical controls for academic ability, race, gender, socioeconomic status, the characteristics of the institution attended, college grades, place of residence, and educational attainment.

Career Mobility and Success

            In our 1991 synthesis, we concluded, albeit cautiously, that a student's major field of study as an undergraduate may have little to do with his or her long-term career mobility in business.  We also pointed out, however, that in the private sector this may well depend on the employing company.  Although not particularly consistent across studies, the results of a very small body of research in the 1990s at least challenge our previous conclusion (Ishida et al., 1997; L. Solnick, 1990; Spilerman & Lunde, 1991).  Spilerman and his colleagues conducted two studies of the factors that influenced job promotion prospects in a single large insurance company.  In the first of these, Spilerman and Lunde (1991) sought to account for promotion within six salary grade intervals and introduced statistical controls for years of education, gender, age, race, and seniority.  In the presence of these controls, employees who were mathematics/science/engineering majors as undergraduates had significantly higher rates of promotion in the middle organizational ranks of the organization than did employees who majored in the humanities or social services.  At the highest organizational ranks of the company below vice president, both business/insurance majors and mathematics/science/engineering majors were significantly more likely to be promoted than were humanities or social sciences majors.

            Somewhat different findings are presented by Ishida, Spilerman, and Su (1997) in a further analysis of the same data.  This difference is probably due to the fact that they operationally defined promotion, not within salary grade intervals, but rather in terms of three major transitions: from clerical to administrative, from administrative to senior management, and from senior management to vice presidential grade.  They also operationally defined major with a different set of categories.  With controls for level of formal education, age, race, sex, seniority, and institutional selectivity, undergraduate social science and economics majors were significantly more likely than humanities majors to be promoted from administrative to senior management ranks.  The probability of promotion from administrative to senior management ranks for both business and science/mathematics majors was not significantly higher than that of humanities majors.  College major had no significant net impact either on the probability of promotion from clerical to administrative ranks or from senior management to vice presidential rank.

            The waters are muddied still further by Solnick's (1990) study of the job success of 370 African-American college graduates employed by a large manufacturing firm.  With controls for such factors as sex, marital status, college grades, timing of bachelor's degree, prior experience, salary grade, and the characteristics of the institution attended, undergraduate major (engineering, science, business or other) had no significant impact on the probability of being promoted within two years after being hired.  However, business majors were advantaged in terms of percentage increases in wages over time.

            It is difficult to form a conclusion about this body of evidence, except to say that, net of other factors, undergraduate college major can play a significant role in some aspects or levels of career mobility and success.  It would appear, however, that the nature and magnitude of this impact depends upon the type of company or firm being considered and its unique cultural norms and values, the particular period of time in one's career, the sector of the company in which individuals with certain undergraduate majors tend to be placed, and the level of promotion or advancement being considered.  As in our previous synthesis, we found little evidence to suggest that undergraduate field of study plays a significant role in promotion to the very highest levels of corporate management or leadership (i.e., vice president or above).

Job Satisfaction

            We uncovered two studies in the literature of the 1990s which estimated the impact of academic major on job satisfaction (Bowen & Bok, 1998; Fricko & Beehr, 1992).  These studies take a very different approach to defining the relationship between academic major and job satisfaction.  Analyzing the 1995 follow-up of the 1976 entering cohort from the College and Beyond data, Bowen and Bok (1998) sought to determine the factors that predicted the likelihood of being "very satisfied" with their job.  Statistical controls were introduced for an extensive array of potentially confounding influences such as race, gender, SAT score, high school rank in class, socioeconomic status, the selectivity of the institution attended, college academic achievement, educational attainment, income, the labor market sector of employment (e.g., profit, nonprofit, self-employed), and marital status.  In the presence of these controls, one's undergraduate academic major (categorized as social science, natural science, engineering, humanities, and other) had only a trivial and statistically nonsignificant impact on being "very satisfied" with one's job.

            Given the extended time period over which the Bowen and Bok (1998) study followed the 1976 entering cohort and the potential for extensive intervening influences, it is not particularly surprising that one's undergraduate academic major exhibited little impact on job satisfaction.  In contrast, Fricko and Beehr (1992) followed up a sample of alumni from a single university who had been out of college for less than five years and had been in their present jobs for between one and two years.  Moreover, instead of assessing the simple net relationship between academic major and job satisfaction, they were interested in how job satisfaction is influenced by the congruence between one's academic major and one's job.  Major-job congruence was measured in two ways.  The first was perceived congruence which was based on responses to the item: "My job is in the same field as my college major," answered on a 7-point scale ranging from "strongly agree" to "strongly disagree."  The second was objective congruence which was based on the degree of match between the person's official major defined in Holland model codes and the person's job defined by the Dictionary of Holland Occupational Codes.  With statistical controls for gender, the gender concentration in the job held (according to the U.S. Bureau of the Census), and salary, both perceived and objective measures of major-job congruence had statistically significant, positive effects on respondents' reported job satisfaction.  In both regression specifications, the positive impact of major-job congruence on job satisfaction was about equal to or greater in magnitude than the positive impact of salary.

            In our 1991 synthesis, we concluded that it was not clear that job fit with one's major was an important determinant of job satisfaction.  Although it is based on a single institution sample and, as far as we know, awaits replication, Fricko and Beehr's (1992) findings suggest that we need to modify our earlier conclusion.

Earnings

            By far, the greatest volume of research on academic major and career attainment focuses on the impact of one’s major field of study on subsequent earnings.  In our present synthesis, we review evidence from 23 individual published and unpublished studies that appeared between 1989 and 2000.  These investigations analyzed data from numerous data sets.  The specific data sets and the studies that employ them were as follows:

            1.   The National Longitudinal Study of the High School Class of 1972-1979 and 1986 follow-ups (Arcidiacono, 1998; Grubb, 1995a, b, 1998, August; James & Alsalam, 1993; James et al., 1989; Knox et al., 1993; Loury, 1997; Loury & Garman, 1995; Rothstein, 1995; Sweetman, 1994a, 1994b).

            2.   The High School and Beyond 1980 cohort followed up in 1986 (Fox, 1993; Hilmer, 2000), and in 1991 (Fitzgerald, 2000).

            3.   The Baccalaureate and Beyond Study of 1992-1993 graduates followed up one year later (S Thomas, 1998) and four years later (S. Thomas, 2000).

            4.   The College and Beyond 1976 cohort followed up in 1995 (Bowen & Bok, 1998).

            5.   The National Center for Education Statistics Surveys of Recent College Graduates: 1985-86 graduates followed up in 1987 (Rumberger & Thomas, 1993); 1989-90 graduates followed up in 1993 (Tsapogas et al., 1994).

            6.   The National Science Foundation New Entrants Survey of 1992 graduates followed up in 1993 (Tsapogas et al., 1994).

            7.   The Cooperative Institutional Research Program data: 1971 freshmen followed up in 1980 (Stoecker & Pascarella, 1991).

            8.   The 1987 and 1990 cohorts of individuals 25-64 years of age from the cross-sectional Survey of Income and Program Participation (Grubb, 1995b, 1997, 1998, August).

            9.   The 1993 follow-up of graduates of the class of 1982 at 20 prestigious colleges and universities (Dowd, 1999).

           10.   A sample of African-American employees of a large manufacturing firm who were hired between 1976 and 1982 (L. Solnick, 1990).

           11.   Two single-institution samples of graduates which estimate starting salary and subsequent salary (Dutt, 1997).

            Drawing conclusions from this body of evidence is complicated by several factors.  First, as might be anticipated, the studies use very idiosyncratic methods for operationally defining undergraduate field of study.  Most use categorical (dummy) variables to indicate different majors (e.g., physical sciences, social sciences, mathematics, humanities, education, and the like).  However, there is only partial consistency in how major is categorized across studies.  Some studies, such as Stoecker and Pascarella (1991), do not even use categories or clusters to define academic field of study, but rather consider it a continuous variable defined by the percent of men in each specific major.  Second, the studies vary substantially in terms of the period of time over which individuals are followed in their careers.  Some studies are concerned with the impact of academic major on starting salary or earnings early in one’s career, while others follow up with individuals ten, or even fifteen, years after college graduation.  Cross-sectional data, such as that from the Survey of Income and Program Participation actually includes information on full-time employees in nearly all age groups.  Finally, the investigations vary to a substantial degree in terms of the statistical controls they are able to introduce for potentially confounding influences.  The typical study, however, introduces controls for such confounding influences as sex, race, socioeconomic background, tested academic ability, precollege aspirations, the selectivity and other characteristics of the undergraduate institution attended, college grades, educational attainment, and measures of work or labor market experience.

            Despite these complications, it is still possible to offer the following generalizations from this body of evidence.  First, consistent with our 1991 conclusions it would appear that undergraduate major field of study has a substantial and statistically significant net impact on earnings that cannot be accounted for by other influences, including the background characteristics of students selecting different majors.  With other factors controlled, there is typically between a 25 to 35% difference in the earnings of individuals who were in different fields of study as undergraduates.  Also consistent with our previous synthesis, the largest earnings premia accrue to majors characterized by a number of traits.  These include a relatively specific and well-defined body of content knowledge and skills, an emphasis on methods of inquiry that require a high level of quantitative and/or scientific skills, a generally close and direct functional link to occupations with relatively high average earnings, in many cases an applied orientation, and a history of being dominated by male students.  Examples of such majors are engineering, business/accounting, several of the physical sciences, mathematics and computer science, and preprofessional majors in health sciences areas such as medicine and dentistry.[30],[31]  This general conclusion does not appear to be seriously affected by differences in the methodological rigor of the studies we reviewed.  Moreover, while the net effect of academic major on earnings appears to be most definitive or pronounced in terms of starting salary or early in one’s career, the general pattern of economic returns to different majors appears to hold later in one’s career.

            Second, there may be an exception to this general pattern for students attending particularly selective, prestigious institutions.  For example, Dowd (1999) analyzed data from the 1993 follow up of 1982 graduates of 20 of the nation’s most selective colleges and universities.  Statistical controls were introduced for ethnicity, region of the country where one held a job, marital status, number of children, and the selectivity of the institution attended.  In the presence of these controls, female graduates from history and political science departments earned about as much as did female graduates of mathematics and physical science departments.  Male history and political science majors actually earned somewhat more than their counterparts with mathematics or physical sciences majors.  Dowd concludes, along with Eide and Waehrer (1998), that in elite institutions liberal arts disciplines may provide the “option value” of potential graduate or professional study.  This functions to enhance longer-term earnings and alter the more typical or representative pattern of major field effects on the economic returns to college.  Bowen and Bok’s (1998) analyses of the College and Beyond sample yielded results for both women and men, but particularly the latter, that are generally consistent with those of Dowd.  The College and Beyond sample they analyzed consisted of a 1995 follow up of individuals who in 1976 enrolled in 28 of the country’s most selective private and public institutions.

            Third, while the vast majority of studies focus on the economic returns to different college majors at the baccalaureate level, there is also a small body of research that estimates the returns to different fields of study at the sub-baccalaureate level.  Most of the evidence is provided by the work of Grubb (1995b; 1997; 1998, August).  Analyzing data from two national data sets (the 1986 follow up of the National Longitudinal Study of the High School Class of 1972 and the 1987 and 1990 cohorts of the Survey of Income and Program Participation), Grubb found that, net of other factors, there were substantial differences in the earnings of individuals with certificates or associate’s degrees in different areas of study.  Generally, for men the economic returns were highest for majors in technical fields (e.g., engineering/computers) and business, while the highest returns accrued to women with majors in business and health.  Unlike the general trend for a bachelor’s degree, however, obtaining an associate’s degree or certificate in some fields of study, such as education, the humanities, and the social sciences for men, often provided little consistent advantage in earnings over a high school diploma.  Conversely, it is clear from Grubb’s extensive analyses that returns to associate’s degrees and to bachelor’s degrees overlap—largely due to an individual’s field of study.  For example, men can generally realize a larger economic premium by obtaining an associate’s degree in engineering, public service, or vocational/technical areas than they can from a bachelor’s degree in the humanities or education.  Similarly, women can generally get a greater earnings return from an associate’s degree in business or health than from a bachelor’s degree in the humanities or education.

            Fourth, a small, but consistent, body of evidence suggests that earnings are enhanced by the extent to which one’s undergraduate major is related to, or congruent with, his or her job.  This finding generally holds at both the baccalaureate and sub-baccalaureate level.  For example, analyzing data from the 1990 cohort of the Survey of Income and Program Participation, Grubb (1997) developed a categorical matching algorithm that links fields of study with Census occupational codes.  He then regressed 1990 earnings on years of formal education completed, plus the matching algorithm indicating if an individual’s major was or was not related to his or her employment.  At the bachelor’s degree level, men in related employment had about a 17% advantage in earnings over men in unrelated employment.  (Here, and in all subsequent estimates, we took the natural antilog, minus one, of the coefficients reported by Grubb and divided the larger by the smaller.)  The corresponding advantage for women in related employment was about 43%.  At the associate’s degree level, men in jobs related to their major had about a 15% advantage over men in unrelated employment, while the advantage for women was about 50%.  At the sub-baccalaureate certificate level, men in related employment demonstrated no earnings advantage over men in unrelated employment; but women in related employment had about a 29% advantage.  Consistent findings are reported in analyses of three national samples by Kolb (1989) and Tsapogas, Cahalan, and Stowe (1994) and in analyses of single institution samples by Callaway, Fuller, and Schoenberger (1996), Dutt (1997), and Fuller and Schoenberger (1991). The findings hold, irrespective of whether the study employs an objective or self-reported (perceptual) measure of major-job congruence or whether the study is predicting starting salary or earnings about five or more years into one’s career. As we pointed out earlier in this chapter, however, part of this finding may be due to the fact that the majors most likely to lead to related employment (e.g., engineering, business, health) are also linked to the highest earnings. Thus, the causal mechanism underlying this finding may be somewhat difficult to ascertain.

Gender Differences in Earnings

            It is clear that one’s undergraduate academic field of study plays a substantial role in determining one’s earnings, particularly in the early stages of one’s career.  It is also clear, from a substantial body of evidence, that there have traditionally been pronounced differences in the pattern of majors populated by men and women (Adelman, 1990, 1991, June, 1998a; Alsalam & Rogers, 1991; Davies & Guppy, 1997; Dowd, 1999; Dutt, 1997; Jacobs, 1996b; Loury, 1997; National Center for Education Statistics, 1997; Nelson & Dixon, 1997; O'Shea, 1989; Sumner & Brown, 1996).  Generally, men tend to be overly represented in fields of study that are closely linked to the highest paying occupations.  These lucrative fields of study include engineering, business, economics, mathematics/statistics, the physical sciences, and the like.  Women, on the other hand, tend to be overrepresented in fields of study that are linked to lower-paying occupations.  These nonlucrative fields of study include the social sciences, humanities, nursing, education, English, journalism, and the like.  Moreover, women are more likely than men to enter nonlucrative fields of study even after such influences as family structure, home environment, age, ethnicity, socioeconomic background, tested ability, and high school curricular track are taken into account (Davies & Guppy, 1997).[32]

            Although it is not unanimous (e.g., Turner & Bowen, 1999), the general view is that this gender segregation in major fields of study has been slowly diminishing in the last three decades (Adelman, 1998; Blau, Ferber, & Winkler, 1998; Eide, 1994; Jacobs, 1995, 1996b).  In turn, there is evidence suggesting that gradual convergence in the gender distribution across major fields of study may explain at least part of the decline in the earnings differential between college-educated men and women during the same period of time (e.g., Blau et al., 1998; Eide, 1994; Light & Ureta, 1990; Loury, 1997).  It is problematic, however, that parity in the distribution of men and women in academic fields of study would totally eliminate gender differences in earnings.  Despite some exceptions (e.g., Paglin & Rufolo, 1990), the weight of evidence is clear in indicating that differences in academic field of study chosen by men and women during college account for part, but not all, of the gender gap in earnings.  When undergraduate academic major is taken into account, the difference in earnings attributable to gender typically becomes smaller (e.g., Sweetman, 1994a, 1994b; S Thomas, 1998; S. Thomas, 2000), but it continues to be statistically significant and often substantial (Arcidiacono, 1998; Callaway et al., 1996; Fox, 1993; Fuller & Schoenberger, 1991; Knox et al., 1993; Rumberger & Thomas, 1993; L. Solnick, 1990; Sweetman, 1994a, 1994b; S Thomas, 1998; S. Thomas, 2000; Tsapogas et al., 1994).  This evidence is consistent with the conclusions from our 1991 synthesis.

Academic Achievement

            In their discussion of the validity and fairness of alternatives to standardized cognitive tests in employment settings, Reilly and Warech (1993) argue that for employers academic achievement (hereafter grades) is a convenient, quantitative summary of a prospective employee’s college performance.  Employers, they assert, believe that grades measure motivation or conscientiousness, as well as cognitive ability.  As such, grades provide a reasonably useful predictor of training success, in most types of jobs, and of actual job performance—particularly in technical fields such as accounting or engineering (Reilly & Warech, 1993).  Indeed, the evidence from research on individual differences suggests that the personal characteristics of cognitive ability and conscientiousness play a major role in career success (Roth & Clarke, 1998; Schmidt, Ones, & Hunter, 1992).  As suggested by Roth and Clarke (1998), substantial correlational evidence links college grades with cognitive ability while other evidence links grades with measures of conscientiousness (e.g., Roth & Clarke, 1998; Schmitt, Ryan, Stierwalt, & Powell, 1995; Wolfe & Johnson, 1995).  Consequently, it is not particularly surprising that the impact of grades on various dimensions of career and economic success has continued to be a major focus of scholars in the decade of the 1990s.  We synthesize this evidence within the following categories:  workforce participation, the link between grades and adult success, occupational status, and earnings.

Workforce Participation

            The evidence on college grades and workforce participation is mixed.  There appears to be little support for the hypothesis that, net of other factors, grades directly influence the long-term likelihood that women will be in the workforce.  For example, when statistical controls were introduced for such factors as precollege educational and occupational aspirations, high school grades, socioeconomic status, institutional selectivity and size, marital status, college major, and educational attainment, Stoecker and Pascarella (1991) found that college grades had no significant effect on women’s likelihood of being full-time in the workforce about nine years after entering college.  Using the same general analytic design, Bowen and Bok (1998) reported similar results.  College grades did not significantly influence a woman’s decision to work when measured about 19 years after entering college.

            Of course, the fact that the Stoecker and Pascarella (1991) and Bowen and Bok (1998) studies considered workforce participation substantially after students entered, and presumably graduated, from college may in part explain the fact that college grades had only a trivial and statistically nonsignificant impact.  Studies that consider workforce participation within one or two years after obtaining a bachelor’s degree report different results.  For example, in analyses of two respective single institution samples that followed up alumni during the first year after graduation, Grayson (1997) and Sagen, Dallam, and Laverty (1997) introduced statistical controls for important student background characteristics and abilities.  In the presence of these controls, college grades significantly and positively influenced the odds of being employed full-time (Grayson, 1997) and being employed in a job appropriate to a bachelor’s degree (Sagen et al., 1997).  Another single-institution study of recent graduates conducted by Williams and Ball (1993) reports findings consistent with those of Grayson (1997), although it is not clear that controls were in effect for potential confounding influences.

The Link Between College Grades and Occupational Success

            One approach to studying the impact of college grades on career and economic success has simply been to estimate the strength of the association between cumulative undergraduate grades and various measures of success in the workplace.  There is no shortage of studies on this topic.  Fortunately, since 1989 there have been a number of quantitative syntheses (or meta-analyses) of this evidence.  Bretz (1989), synthesizing 39 studies, Dye and Reck (1989), synthesizing 72 studies, and Roth, BeVier, Switzer, and Schippman (1996), synthesizing 49 studies, have all estimated the link between undergraduate grades and various measures of job success.  The measures of job success included supervisor’s ratings of job performance, measures of output/productivity, salary, training success, and the like.  Weighting the overall findings of each meta-analysis by the number of studies reviewed, we estimate that the average correlation between cumulative undergraduate grades and job success/performance was about .16.  In other words, across all studies undergraduate grades accounted for about 2.6% of the variance in job success (i.e., .162).  Overall, this is not a particularly strong association.  However, grades are a better predictor of performance as the links between education and jobs get closer.  Examples would include such fields as business and engineering.  Moreover, Roth, BeVier, Switzer, and Schippman (1996) also found that when the simple correlation between grades and job success was corrected for such factors as the reliability and range restriction of grades and job performance measures, the size of the correlation increased to between .23 and .36.

            More recently, Roth and Clarke (1998) conducted a meta-analysis of 75 studies focusing on the relationship between undergraduate grades and earnings or salary.  They found a simple correlation of .14 between grades and starting salary, which rose to .22 when the correlation was corrected for range restriction and the unreliability of grades.  College grades had a simple correlation of .17 with currently salary, which rose to .26 when the correlation was corrected for range restriction and the unreliability of grades.  The correlation between grades and salary growth was negligible.  Using the correlations derived by Roth and Clarke from their meta-analysis, one would conclude that undergraduate college grades are associated with between 2 and 4.8% of the variance in starting salary (i.e., .142 and .222) and 2.9 and 6.8% of the variance in current salary.  The lower estimates are quite consistent with the conclusions from our previous synthesis.  The higher estimates are somewhat larger than our 1991 estimates, but that is probably due to the corrections for restricted range and unreliability of grades included in Roth and Clarke’s work.

            It is also worth noting that, consistent with the meta-analysis results concerning grades and job success, Roth and Clarke (1998) also found that in fields where there was a stronger connection between education and job skills the correlation between grades and salary was larger.  For example, in engineering the correlations were in the .35 to .45 range.

            Of course, there is a significant problem with this body of evidence.  Namely, the simple correlation between college grades and various measures of job performance and success is likely confounded by other factors.  Thus, only part of the association may be causal.  Another body of research has attempted to estimate the actual net or causal impact of college grades, career, and economic success.  We turn now to a summary of that evidence.

Job Satisfaction and Job Mobility

            As with our previous synthesis, we found little evidence to suggest that undergraduate grades are causally linked with job satisfaction.  Both Bretz (1989), analyzing a single institution sample, and Bowen and Bok