1 nfl n n L b r r t t f fr n r n ll r d t : N t n l t d T rr ll L. tr h rn The Journal of Higher Education, Volume 79, Number 1, January/February 2008, pp (Article) P bl h d b Th h t t n v r t Pr DOI: /jhe For additional information about this article Access provided by Penn State Univ Libraries (17 Jun :53 GMT)
2 Terrell L. Strayhorn Influences on Labor Market Outcomes of African American College Graduates: A National Study Much of the research on the effects of college suggests that earning a bachelor s degree significantly influences one s economic success and labor market outcomes such as earnings, job security, and prestige of one s occupation (Ehrenberg & Rothstein, 1994). For example, several studies provide evidence to support the belief that college graduates earn higher annual salaries than do high school graduates (Pascarella & Terenzini, 1991, 2005; Smart, 1986; Smart & Pascarella, 1986) and are less likely to face periods of unemployment. A report from the U.S. Department of Education (2000) indicates that bachelor s degree (BA) recipients earned between 1.4 and 1.8 times more than those with only a high school diploma or its equivalent in Approximately 10 years later, BA recipients earned between 1.6 and 2 times more than those who graduated from high school only. Taken together, these results suggest that receiving a BA yields a substantial increase in earnings over one s lifetime. In fact, reports in the popular press suggest that college graduates earn approximately $1 million dollars more than nongraduates over their lifetime (Day & Newburger, 2002). However, a segment of this body of research provides evidence of differences in the labor market outcomes of African American college graduates (Allen, 1992; Constantine, 1995; Sagen, Dallam, & Laverty, The author wishes to thank Dr. Elchanan Cohn and two anonymous reviewers for their guidance on this manuscript. Terrell L. Strayhorn is Assistant Professor of Higher Education and Special Assistant to the Provost at The University of Tennessee Knoxville. The Journal of Higher Education, Vol. 79, No. 1 (January/February 2008) Copyright 2008 by The Ohio State University
3 Outcomes of Black College Graduates ; Thomas, 2000). The weight of the evidence suggests that African American college graduates are at a disadvantage with respect to postgraduate earnings and occupational status. Despite some progress, still today the Black unemployment rate is more than two times that of White Americans, and Black families earn only 58% as much income as White families. Perhaps an even more dramatic depiction of the current situation is reflected in national net worth comparisons: In 2001, the typical Black household had a net worth of just $19,000 (including home equity) compared with $121,000 for Whites (Muhammad, Davis, Lui, & Leondar- Wright, 2004, p. 1). Although the evidence is quite clear that African Americans face significant disadvantages with respect to labor market outcomes, it is less clear just why this is the case. That is, the causal mechanism underlying this disadvantage is difficult to ascertain but an important and necessary piece of the puzzle. As a result, some studies have examined the influence of race on economic or labor market outcomes (Hoffman, 1984; Pascarella, Smart, & Stoecker, 1989; Phelan & Phelan, 1983). For example, researchers have shown that African Americans, as a whole, earn lower annual salaries than any other racial group (National Center for Education Statistics [NCES], 2001). Moreover, other studies provide evidence that African Americans reported lower job satisfaction than did their White counterparts (Phelan & Phelan, 1983) and reported lower scores on job satisfaction than did other racial groups (Mau & Kopischke, 2001). Perhaps the largest single body of research on the labor market outcomes of African American college graduates concerns the impact of attending a historically Black college or university (HBCU) (Allen, 1992; Constantine, 1994, 1995; Ehrenberg & Rothstein, 1994; Fitzgerald, 2000; London, 1998; Solnick, 1990; Thomas, 2000). Findings concerning the net impact of graduating from an HBCU on African American college graduates economic success are mixed. For example, Ehrenberg and Rothstein analyzed national data and found that attending an HBCU had a statistically nonsignificant effect on subsequent occupational status and earnings, controlling for gender, SAT scores, high school rank, educational attainment, and a number of other confounding influences. On the other hand, Constantine studied African Americans at four-year institutions and found that attendance at an HBCU versus a predominately White institution (PWI) had a statistically significant positive effect on graduates earnings, controlling for a battery of individual level characteristics such as high school achievement and gender. More evidence is needed to substantiate the net impact of HBCU graduation on the post-ba labor market outcomes of African American college graduates. This is the gap addressed by this study.
4 30 The Journal of Higher Education In sum, several major themes were found in the literature. First, college graduates reap significant economic benefits or private returns on their investments in higher education (Day & Newburger, 2002; Ehrenberg & Rothstein, 1994; NCES, 2000). However, African American college graduates still face significant disadvantages with respect to post-ba earnings and occupational status. A limited number of studies also examined job satisfaction for African American college graduates (Mau & Kopischke, 2001; Phelan & Phelan, 1983) and found similar conclusions. While the weight of evidence provides clear and compelling information about the economic disparities of African American college graduates, much less is known about the underlying causal mechanism and those factors that influence their post-ba labor market outcomes. One possible factor that has received a relatively significant amount of research attention is the impact of graduating from an HBCU. The weight of evidence is inconsistent in suggesting that, net of other influences, graduating from an HBCU has a significant, positive impact on subsequent earnings and other labor market outcomes. Some studies support this conclusion (Constantine, 1994, 1995), while others provide little to no support (Ehrenberg & Rothstein, 1994; London, 1998). Indeed, some research has shown that HBCU graduates face a statistically significant disadvantage in subsequent earnings (Thomas, 2000). Given these equivocal findings, estimates of the net impact of HBCU graduation on post-ba labor market outcomes of African American college graduates are obscured and in need of additional empirical testing. Thus, the purpose of this study was to estimate the effects of factors that influence the post-ba earnings, occupational status, and job satisfaction of African American college graduates. These factors include background traits, precollege characteristics, institutional characteristics, college experiences, and post-ba experiences. Specifically, this analysis centered on the relationship between graduating from an HBCU and one s subsequent economic outcomes (e.g., earnings, occupational status, and job satisfaction). Using data from the NCES Baccalaureate and Beyond Longitudinal Survey (B&B:1993/1997; Green, Myers, Veldman, & Pedlow, 1999), this study sought to determine the impact of students traits, postsecondary experiences, and post-ba experiences on labor market outcomes of Black college graduates. Specifically, the following research questions guided this investigation: 1. Do HBCU graduates and non-hbcu graduates differ on three measures of labor market outcomes (e.g., salary, occupational status, and job satisfaction)? 2. What is the net effect of attending an HBCU on the postbaccalaureate earnings of African American college graduates?
5 Outcomes of Black College Graduates What is the net effect of attending an HBCU on the occupational status of African American college graduates? 4. What is the net effect of attending an HBCU on the job satisfaction level of African American college graduates? 5. Are the effects of HBCU attendance conditional on the basis of gender? This analysis differs from prior research studies in several ways. First, prior research consists largely of single-institution or small student samples (Constantine, 1994, 1995); samples from a single employment sector (Solnick, 1990); and even single states (Johnson, 1982). This analysis was based on nationally representative data drawn from a large-scale survey of students from multiple institutions and across various academic majors. Second, previous studies tend to examine the outcomes of college using simple correlational or flat analytical techniques without statistical controls. The absence of statistical controls proves problematic (Keith, 2006), thus potentially biasing the estimates of the effects upward by not accounting for the confounding influences of other independent variables. In this analysis, an extensive array of statistical controls for potentially confounding variables was included to isolate the net impact of HBCU graduation on labor market outcomes. Finally, this investigation employed a theoretical framework to establish relationships among the variables and to guide the selection of variables (and proxies) included in the statistical model. In part, this allows one to test the power of an expanded, hierarchical model in explaining labor market outcomes of recent African American college graduates. Theoretical Framework This study of the effects of college on African American students labor market outcomes was guided by a number of theoretical explanations. First, this study was informed by human capital theory (Becker, 1993; Schultz, 1971). Human capital suggests that an individual makes investments in education or training, thereby gaining additional skills and knowledge that are often associated with increased likelihood of occupational attainment and economic success. Broadly conceived, human capital refers to the information, knowledge, skills, and abilities of an individual that can be exchanged in the labor market for returns such as salary, financial rewards, and jobs. In short, the more education an individual attains, the more human capital one accumulates and then the more an individual gains with respect to outcomes.
6 32 The Journal of Higher Education Occupational attainment research has also shown that other factors influence economic outcomes of college graduates. College major has been found to play a significant role in predicting after-college outcomes of graduates. Graduates in science and engineering fields earn higher salaries than those who major in social sciences and humanities. Other findings support this conclusion and highlight the way in which college major impacts post-ba earnings and other measures of labor market success, including job satisfaction (Bisconti & Solomon, 1977; Johnson, 1982; Pascarella & Terenzini, 2005; Smart, 1986). These relationships are also substantiated by findings from studies on the effects of college attendance on socioeconomic attainment (Astin, 1977, 1993; Pascarella & Terenzini, 1991; Terenzini & Wright, 1987). Continuing with this line of thought, grade point average (GPA) was expected to be related to economic outcomes of college graduates such as annual earnings. Findings suggest that college grades have a positive effect on income and that the effect is stronger for African Americans than for Whites (Pascarella & Smart, 1990). Other researchers have studied the effect of background traits and grades on labor market outcomes and found that grades also influence post-ba job satisfaction (Johnson, 1982). This study employed an integrated model that expands traditional econometric models that are typically applied in economic analysis by including measures of social and cultural capital. Like human capital, social and cultural capital are resources that can be invested to enhance profitability (Bourdieu & Passeron, 1977), increase productivity (Coleman, 1988), and facilitate upward mobility (DiMaggio & Mohr, 1985; Lamont & Lareau, 1988). Social capital takes the form of information-sharing networks as well as social norms, values, and expected behaviors (Coleman, 1988). Social capital also refers to the way in which those connections are maintained (Morrow, 1999). Cultural capital, on the other hand, is the system of beliefs, tastes, and preferences derived from one s parents (or guardians) that ultimately define an individual s class status (Bourdieu & Passeron, 1977; McDonough, 1997). One way social and cultural capital may influence one s economic and labor market outcomes is through the provision of knowledge and information about college, jobs, and career options (Bourdieu & Passeron, 1977; McDonough, 1997). In this model, proxies for the availability of information about college, job, and career choices include type of high school attended, family income, and college selectivity. Previous studies set the precedent for using such factors (McPherson & Winston, 1993; Perna, 1998, 2004; Trusheim & Crouse, 1981; Zhang, 2005). Also, I control for geographic region of undergraduate college because regional
7 Outcomes of Black College Graduates 33 differences may reflect variations in salary and the presence of an HBCU (Constantine, 2000; McDonough, Antonio, & Trent, 1995). Social and cultural capital may also refer to one s values and preferences for education, such as one s values about obtaining a college degree (DiMaggio & Mohr, 1985; McDonough, 1997). In this analysis, proxies for the value placed on education include students educational expectations and parental educational attainment. Given previous findings that suggest differences between the influence of mother s and father s educational background for African Americans (Maple & Stage, 1991; Strayhorn, McCall, & Jennings, 2006), two separate measures were included to reflect the educational attainment of each parent. Including such measures allowed me to test whether social and cultural capital variables increase the predictive power of a typical econometric outcome model. In sum, theoretical explanations and empirical research findings provide support for the influence of earning a BA degree on labor market outcomes such as annual earnings and job satisfaction (Becker, 1993; Bisconti & Solomon, 1977; Geske, 1996; NCES, 2000; Smart, 1986). Findings also provide evidence to suggest that this relationship is mediated by factors such as college major (Bisconti & Solomon, 1977; Johnson, 1982; Pascarella & Terenzini, 2005; Smart, 1986), college grades (Johnson, 1982; Pascarella & Smart, 1990), and race (Hoffman, 1984; Mau & Kopischke, 2001; Pascarella et al., 1989; Phelan & Phelan, 1983). Finally, sociocultural explanations also posit how background and environmental factors relate to economic outcomes such as salary and occupational status (Coleman, 1988; DiMaggio & Mohr, 1985; Lamont & Lareau, 1988; Paulsen, 2001). Prior research has shown that expanded econometric models that include measures of human, social, and cultural capital are improved over traditional economic models when explaining college student decisions and outcomes such as enrollment in college (Perna, 2000), pursuit of graduate study (Perna, 2004), and graduate student persistence (Strayhorn, 2005). Therefore, it seems plausible that an expanded model may also be more useful when studying post-ba labor market outcomes. Thus, another purpose of this study was to explore this hypothesis: Do measures of human, social, and cultural capital add to the power of statistical models to explain variance in post-ba outcomes? Method Data Source Data were drawn from the National Center for Education Statistics Baccalaureate & Beyond Longitudinal Study (B&B:93/97). The B&B
8 34 The Journal of Higher Education study follows baccalaureate degree completers over time to provide information on work experiences after college and post-ba outcomes such as earnings. Using NPSAS:93 as the base year, the B&B:93/97 Longitudinal Study follows baccalaureate degree completers beyond their undergraduate graduation (U. S. Department of Education, 1999). This is particularly useful for studying the effect of college on post-ba labor market outcomes such as annual earnings. In addition, given the maximum economic return is associated with graduating from college (Murphy & Welch, 1989; Rupert, Schweitzer, Serverance-Lossin, & Turner, 1996; Turner & Bowen, 1990), this data source was most appropriate as it provides information on a national sample of college graduates. The follow-up surveys provide a unique opportunity to gather information concerning delayed entry into graduate education, graduate school aspirations, persistence, and the interaction between work and education experiences beyond obtaining a bachelor s degree (U. S. Department of Education, 1999). The first-year follow-up (B&B:93/94) surveys BA recipients one year after receiving their college degree, while the second follow-up (B&B:93/97) elicits information about participants four to five years after graduation. These data were deemed appropriate for this investigation and have been used in previous studies to explore the decision to enroll in college (Perna, 2000, 2004) and graduate student persistence (Strayhorn, 2005). From the NPSAS:93 sampling criteria, 16,316 baccalaureate degree recipients were identified. All those who completed the NPSAS interview and for which NPSAS parent data were available were retained. The total sample included 11,192 cases that were retained for future rounds, including the second follow-up. The present study used data drawn from the B&B:93/97 second follow-up study. For the second follow-up, the total sample consisted of 9,274 respondents, 83% of the original sample. For this analysis, the sample was restricted to African American students only. The weighted sample size was 71,831. The majority were female (67%), and 33% graduated from an HBCU while 67% did not. Table 1 presents additional information to describe the sample. Variables The dependent variables in this study are measures of labor market success namely, annual earnings, occupational status attainment, and job satisfaction. Specifically, one dependent variable measured the annual salary (in dollars) of recent college graduates as reported on the B&B survey. Occupational status attainment (dependent variable 2) was measured by converting each individual s occupational code to a measure of occupational status attainment as defined by Duncan (1961) and later revised
9 Outcomes of Black College Graduates 35 TABLE 1 Description of Sample Characteristic/Variable % Father s Educational Attainment Not HS graduate or equivalent 8.0 HS graduate or equivalent 30.2 Some postsecondary, less than 2 years years of postsecondary, less than BA 13.6 Bachelor s degree 21.0 Advanced degree 18.6 Mother s Educational Attainment Not HS graduate or equivalent 6.4 HS graduate or equivalent 33.6 Some postsecondary, less than 2 years years of postsecondary, less than BA 8.1 Bachelor s degree 17.0 Advanced degree 10.1 Gender Male 33.3 Female 66.7 HBCU Graduate No 67.0 Yes 33.0 Graduate School Enrollment No 69.4 Yes 30.6 NOTE: HS = high school. BA = bachelor s degree. HBCU = historically Black college or university. by Featherman and Stevens (1982). That is, each occupational code was assigned a socioeconomic index based on extensive research on occupational status (see Featherman & Stevens, 1982, for a full discussion of the socioeconomic index). These variables are consistent with techniques used in previous studies (Ehrenberg & Rothstein, 1994; Lin & Vogt, 1996; Smart, 1986; Trusheim & Crouse, 1981). For the purposes of this study, job satisfaction (dependent variable 3) was defined as the degree of pleasure or happiness derived by employees from their work, work relations, and work-related factors such as salary, fringe benefits, working conditions, opportunity for advancement, and job security, to name a few (Fisher, 2000; Mau & Kopischke, 2001; Price & Mueller, 1986). Theoretically speaking, job satisfaction is based on the degree of congruence between an individual s skills and aspirations and the perceived or actual nature of the job (Bretz & Judge, 1994). Job satisfaction was measured using nine variables from the B&B:93/97 data-
10 36 The Journal of Higher Education base. Similar variables were used in previous research and were deemed appropriate for the current analysis (Mau & Kopischke, 2001). The independent variables consist of five sets of predictors. The first set includes background traits and precollege characteristics. These include race, gender, age, family income, mother s educational attainment, father s educational attainment, type of high school attended, precollege ability as measured by college entrance exam scores, and educational aspirations. Educational aspirations were measured using four categories ranging from less than BA to advanced degree. Parental educational attainment was measured by six categories: less than high school; high school; some postsecondary education, less than BA; bachelor s degree; and advanced degree. The second and third set of predictors included institutional characteristics and academic factors, respectively. Institutional characteristics were measured by whether one graduated from an HBCU, college selectivity defined as the mean value of SAT/ACT scores, and institutional control. Academic variables included college GPA, attained associate s degree, and major. Major was operationalized using a set of four dichotomous variables indicating whether one s major was classified as specialized hard, specialized soft, broad professional, or general liberal arts. This conceptualization was also used in Sagen, Dallam, and Laverty s 1997 study. Finally, nonacademic experiences and post-ba experiences were included in the model. Nonacademic experiences refer to the hours worked per week, while post-ba experiences include participation in graduate education and marital status. Precedent for using these variables to estimate the net impacts of college attendance on student-level outcomes was set in previous studies (Ehrenberg & Rothstein, 1994; Lin & Vogt, 1996; Pascarella & Smart, 1990). Data Analysis Several analytical procedures were used to investigate the research questions. First, descriptive statistics were computed to characterize the sample and to distinguish those who graduated from HBCUs from those who did not. Independent sample t-tests were used to determine differences between these groups on selected background and precollege characteristics, institutional characteristics, and experiences. Finally, hierarchical linear regression techniques were used to measure the influence of such factors on three measures of labor market outcomes namely, annual earnings, occupational status attainment, and job satisfaction. Independent variables were entered into the model proceeding from precollege and background traits, to college experiences (academic and non-academic) and institutional factors, to post-ba experiences. The independent
11 Outcomes of Black College Graduates 37 variable of interest, whether a student graduated from an HBCU, was entered in the last and final model. This statistical design permitted the use of a rigorous set of statistical controls and isolated the net effect of individual sets of predictors on the dependent variable(s) under study. Weighting and Technical Issues While the instruments used for both the NPSAS and the B&B surveys were found to be reliable through field testing and follow-up studies, adjustments must be made to compensate for unequal probability of selection into the B&B sample and to adjust for non-response (U.S. Department of Education, 1999, p. 108). Due to the complex sampling design, appropriate sampling weights must be applied when approximating the population of the bachelor s degree recipients in the longitudinal sample. The B&B:93/97 panel weight is appropriate for this purpose and was applied to provide national probability estimates adjusted for differential rates of selection and nonresponse. To minimize the influence of sample sizes on standard errors while also correcting for the oversampling of some groups, each case is weighted by the panel weight divided by the average weight for the sample [the relative weight] (Perna, 2004, p. 492). All statistical analyses were conducted using AM software (version beta) provided by the American Institutes of Research (2002), which is appropriate for use with weighted data from complex samples. In addition, due to the nested nature of these data, a more rigorous threshold of statistical significance was used to interpret the results where possible (Thomas & Heck, 2001). Despite these adjustments, there are several limitations that should be discussed before presenting the findings from this analysis. Limitations Missing Data Some analyses in this study are limited by the magnitude of missing data. Variables with the largest share of missing data are those pertaining to family income, salary, and age, though all variables in the study were missing less than 10% of cases. In some cases, listwise deletion would reduce the analytic sample significantly and possibly result in a sample that is not representative of the population of bachelor s degree recipients. While researchers disagree about the minimum number of cases that is required per independent variable, most generally agree that larger samples will generate more stable parameter estimates and more accurate χ 2
12 38 The Journal of Higher Education distributions (Peng, So, Stage, & St. John, 2002). To avoid the substantial reduction in sample size that would occur during listwise deletion of missing data and to account for the tendency of cases to be missing data for more than one independent variable, I took several steps to reduce the number of missing cases (Cohen & Cohen, 1983). First, mean scores were imputed for cases that were missing data on continuous independent variables. While these data were imputed to minimize the effects of missing data, this procedure may result in an underestimation of standard errors by 10 20% and increase the chances of making a type-1 error. Therefore, a more rigorous threshold of statistical significance was used when interpreting such results. Some cases were missing data on scale variables. In this case, I used trend equations (Thomas & Heck, 2001) to impute values for the missing cases. Trend equations act much like regression equations and predict missing values using data provided on valid cases in the sample. Predicted values were imputed for all missing cases on scale items, except when missing values constituted no more than 1% of all cases. It is important to note that imputation of mean values in place of missing observations was used only for continuous independent variables, while trend calculations were used to impute values for missing observations on scale items. Missing cases for the dependent variables were excluded from the analysis, as recommended by others (Galloway, 2004; Perna, 2004). Results Descriptive statistics suggest that the sample of 1993 African American bachelor s degree recipients were majority female (67%), and the average age at the time of graduation was years (SD = 7.82). Black graduates average SAT scores (M = ; SD = ) reflect the national average for African Americans at that time (College Board, n.d.). For those who did not take the SAT, average ACT scores were computed (M = 20.89; SD = 3.51). On average, participants worked hours per week while enrolled (SD = 15.05). Results suggest that the sample is sufficiently representative of the population. Table 2 presents means and standard deviations for all independent and dependent variables included in this analysis. Differences in Earnings, Occupational Status Attainment, and Job Satisfaction of HBCU and non-hbcu Graduates An independent sample t test was conducted to determine differences between HBCU and non-hbcu graduates with respect to annual earn-
13 Outcomes of Black College Graduates 39 TABLE 2 Mean and Standard Deviations of Independent and Dependent Variables Independent Variables M SD Gender Age Family income 39, , Mom s level of education Dad s level of education Type of high school Marital status Education aspirations GPA Hours worked ACT Score SAT Score Control Associate s degree Attend graduate school Attend HBCU Annual salary $30, , SEI Satisfaction Weighted N 71,831 NOTE: GPA = grade point average. HBCU = historically Black college or university. SEI = socioeconomic index. ings. HBCU and non-hbcu graduates differed significantly in terms of annual earnings, t(384.28) = 3.36, p < That is, HBCU graduates reported lower salaries (M = 27,910; SD = 15,144) than did their counterparts who graduated from non-hbcu institutions (M = 32,317, SD = 15,145). Independent sample t tests were conducted to test for differences between HBCU and non-hbcu graduates in their post-ba occupational status attainment level (as measured by the socioeconomic index, or SEI) and their level of job satisfaction. The tests were not significant, t(478.19) = 1.76, p = 0.07 and t(421.89) =.11, p = 0.91, respectively. Though HBCU graduates rank higher with respect to SEI (M = 60.62, SD = 21.38) than do non-hbcu graduates (M = 57.40, SD = 23.47), the difference does not reach the level of statistical significance. Even smaller differences are observed for job satisfaction. Table 3 presents a summary of these findings.
14 40 The Journal of Higher Education TABLE 3 Differences between HBCU and non-hbcu Graduates on Selected Variables Variable/Group n a M SD t ACT score Non-HBCU 47, * HBCU 23, SAT score Non-HBCU 47, HBCU 23, Age Non-HBCU 47, HBCU 23, Aspirations Non-HBCU 47, HBCU 23, Satisfaction Non-HBCU 47, HBCU 23, Salary Non-HBCU 47,863 $27, * HBCU 23,968 $32, NOTE: HBCU = historically Black college or university. a Weighted sample sizes are shown in table; adjusted weighted sample sizes were used to conduct analyses. * p <.01 Relationship of Factors with Earnings Exploratory correlation analyses reveal a number of important relationships between the independent and dependent variables. Still, correlation results suggest low to modest relationships. On the one hand, this indicates that multicollinearity is not a problem for this research investigation. On the other, it shows that variables are loosely related and may not be sufficiently related to explain a significant proportion of variance. Table 4 presents a summary of the correlation analysis. Hierarchical multiple regression techniques were used to investigate the relationship between background and precollege characteristics, college variables, post-ba experiences, and annual earnings. That is, a sequential multiple regression was ordered in such a way as to examine the relationship between all of the independent factors (including measures of human, social, and cultural capital) and a measure of labor market outcome, annual earnings. The regression model that included only background and precollege variables (step 1) was significant, F(8,441) = 3.801, p < The sample correlation coefficient was 0.25, indicating that approxi-
15 TABLE 4 Correlations among Selected Independent and Dependent Variables Gender 2. Age Family income 0.11* 0.10* 4. Mom s education * 0.16* 5. Dad education 0.11* 0.21* 0.15* 0.49* 6. GPA Post-BA enrollment * 8. ACT score * * SAT score * 0.16* * Associate s degree * * 0.14* * 11. Annual salary 0.12* 0.17* * 0.12* SEI 0.20* * 13. Job satisfaction 0.10* * * 0.03 NOTE: GPA = grade point average. BA = bachelor s degree. SEI = socioeconomic index. * p < 0.05.
16 42 The Journal of Higher Education mately 6% of the variance in annual earnings can be accounted for by the linear combination of background and precollege measures. Based on these results, background and precollege characteristics appear to be significant, albeit modest, predictors of earnings. After adding the college experiences factors and post-ba variables, including both social and cultural capital measures, the regression model was found to be significant again, F(18,431) = 2.626, p < The sample correlation coefficient was 0.31, indicating that approximately 9% of the variance in earnings can be explained by the linear combination of independent and control variables. Based on these findings, these factors appear to be significant predictors of annual earnings, as college experiences and post-ba experiences add significantly to the power of the model, R 2 = The final hierarchical model consisted of all independent factors that were entered previously in order to test the relationship between HBCU attendance and post-ba annual earnings. The final model was found to be significant overall, F(19,430) = 2.764, p < The sample correlation coefficient was 0.33, indicating that 11% of the variance in earnings can be accounted for by the combination of all predictor variables. Model change statistics indicate that the final model is a significant improvement over the previous models, R 2 = 0.01, F (1, 430) = 4.832, p < Finally, results suggest that several independent variables are significant predictors of the criterion variable, annual earnings. Relative beta weight comparisons suggest that gender, age, hours worked while enrolled, and HBCU attendance have the strongest significant influence on earnings. Results from the final model are reported in Table 5. Relationship of Factors with Occupational Status Attainment or Socioeconomic Index (SEI) Hierarchical multiple regression analyses were conducted to evaluate the relationship between all independent and control variables included in the model and another labor market outcome, occupational status attainment as measured by Duncan s SEI. The regression model was significant, F(18, 539) = 3.521, p < The sample correlation coefficient was 0.32, indicating that approximately 10.5% of the variance in the individual s occupational status attainment level can be accounted for by the combination of independent variables. A second analysis was conducted to evaluate whether the indicator for graduating from an HBCU predicted one s occupational status attainment level over and above the previous model including background, precollege, college, and post-ba experiences. Adding the HBCU variable accounted for a statistically significant proportion of the SEI variance after
17 Outcomes of Black College Graduates 43 TABLE 5 Summary of Model Predicting Earnings from Background, Precollege, College, and Related Variables Variable B SE B β t p (Constant) Gender Age Dad s education Mother s education Marital status High school type Family SES Educational aspirations College GPA Hours worked ACT score SAT score Institutional control Associate s degree Specialized hard major Broad professional major General liberal arts major Attend graduate school Graduate from HBCU R 0.33 R NOTE: SES = socioeconomic status. GPA = grade point average. HBCU = historically Black college or university. controlling for the effects of all previously entered variables in the model, R 2 = 0.007, F(1, 538) = 4.300, p < The sample correlation coefficient was 0.34, indicating that approximately 12% of the variance in SEI scores can be accounted for by the variables in the model. Results suggest that one s aspirations, academic major, attending graduate school, and graduating from an HBCU are significant predictors of occupational status. Results from the final regression model are reported in Table 6. Relationship of Factors with Job Satisfaction Hierarchical or sequential multiple regression analyses were conducted to evaluate the relationship between background characteristics, precollege and college variables, post-ba experiences, and the level of job satisfaction reported by participants. The regression equation was significant, F(18, 540) = 2.331, p < The sample multiple correlation coefficient
18 44 The Journal of Higher Education TABLE 6 Summary of Model Predicting Occupational Status Attainment (SEI) from Background, Precollege, College, and Related Variables Variable B SE B β t p (Constant) Gender Age Dad s education Mother s education Marital status High school type Family SES Educational aspirations College GPA Hours worked ACT score SAT score Institutional control Associate s degree Specialized hard major Broad professional major General liberal arts major Attend graduate school Graduate from HBCU R 0.34 R NOTE: SES = socioeconomic status. GPA = grade point average. HBCU = historically Black college or university. was 0.27, indicating that approximately 7% of the variance in job satisfaction can be accounted for by the combination of independent factors. A second analysis was conducted to estimate the net impact of graduating from an HBCU on one s job satisfaction level. Adding the HBCU variable to the model did not add significantly to the power of the model already containing a number of control variables and measures of social and cultural capital. That is, graduating from an HBCU had a statistically nonsignificant net effect on job satisfaction, F(1,537) =.210, p >.05. Findings suggest that gender, marital status, and college GPA have a statistically significant relationship with job satisfaction for African American college graduates. Results are summarized in Table 7. Conditional Effects of HBCU Attendance To test for conditional effects of HBCU attendance on the basis of gender, a cross-product term was added to each statistical model. Results suggest that HBCU attendance does not have differential effects on salary,
19 Outcomes of Black College Graduates 45 TABLE 7 Summary of Model Predicting Job Satisfaction from Background, Precollege, College, and Related Variables Variable B SE B β t p (Constant) Gender Age Dad s education Mother s education Marital status High school type Family SES Educational aspirations College GPA Hours worked ACT score SAT score Institutional control Associate s degree Specialized hard major Broad professional major General liberal arts major Attend graduate school Graduate from HBCU R 0.27 R NOTE: SES = socioeconomic status. GPA = grade point average. HBCU = historically Black college or university. occupational status, or job satisfaction depending on the sex of the student. That is, the addition of interaction terms to the models did not lead to a statistically significant increase in the model s parameters. Therefore, these results will not be explicated further. According to tolerance statistics, multicollinearity was not a problem for this investigation as the correlations between the independent and dependent variables are moderate to trivial and largely statistically nonsignificant. Moreover, correlations among independent variables were not a cause for concern. Discussion This study employed a hierarchical design with statistical controls for potentially confounding characteristics to estimate the net impact of attending an HBCU on three measures of labor market outcomes using a
20 46 The Journal of Higher Education national sample of African American college graduates. Specifically, this longitudinal analysis examined the influence of attending an HBCU on African American graduates earnings, occupational status or socioeconomic index, and job satisfaction after college. Findings suggest a number of important conclusions. Overall, HBCU graduates and non-hbcu graduates differed significantly on post-ba annual earnings but did not differ in terms of occupational status and job satisfaction. Still, a number of other important relationships should be highlighted. Differences in Earnings of HBCU and non-hbcu Graduates Attending an HBCU was associated with lower levels of annual salary for African American graduates. Such results are consistent with findings reported by Ehrenberg and Rothstein (1994), Thomas (2000), and Fitzgerald (2000). However, they challenge conclusions drawn in Constantine s (1995) study that suggest attendance at an HBCU may exert a positive influence on subsequent wages. While the results of this study present compelling evidence of the impact of HBCU attendance on annual earnings, far less is revealed about the causal mechanism underlying this phenomenon. On the other hand, the results of this study suggest that HBCU attendance may be part of the causal mechanism underlying differences in earnings between African Americans and other racial/ethnic groups (see Phelan & Phelan, 1983; NCES, 2001). By including African Americans only (who represent the largest proportion of HBCU students), this study sought to isolate the true, net effect of HBCU attendance on earnings and to advance this line of inquiry by adjusting the estimates of effects downward by accounting for potentially confounding variables. These findings are important for a number of constituent groups in higher education. Families and students should consider this evidence when making college choices. Yet, caution should be exercised when interpreting the finding that relates to the impact of HBCU attendance on earnings. Previous research provides compelling evidence of the positive effects of attending an HBCU on outcomes for African American students, such as racial ideology (Cokley, 1999), racial identity (McCowen & Alston, 1998), and even racial uplift (Brown & Freeman, 2002; Hirt, Strayhorn, Amelink, & Bennett, 2006). Though evidence about the impact of attending an HBCU on economic success is inconsistent (Pascarella & Terenzini, 2005), prior studies suggest that attending HBCUs has a positive net impact on cognitive and affective outcomes such as knowledge acquisition, intellectual development, academic and social self-concepts (Berger & Milem, 2000), and persistence for African American collegians. Indeed, results from the present study suggest that HBCU attendance has
21 Outcomes of Black College Graduates 47 a negative net impact on future earnings and may provide evidence of employers preferences for non-hbcu graduates rather than an actual negative effect that HBCUs confer upon their students. The research literature provides rather consistent and compelling information about the nurturing environments that Black institutions engender (Allen, 1992; Bonner & Bailey, 2006; Hirt et al., 2006). Of course, there is an obvious alternative hypothesis to explain the differences found relative to the effects of college on earnings. Prior reports indicate that HBCUs tend to offer degrees in some areas (i.e., humanities and social sciences) more than in others (i.e., engineering, medicine, business). To the extent that the effect of HBCU attendance on earnings is related to one s academic major, there may be less cause for concern about employers perceptions. Still because the B&B data do not contain additional information on major offerings of schools (particularly HBCUs) and specific information about coursework, we cannot determine the extent to which such factors may have accounted for the differences observed in this study. Differences in Occupational Status of HBCU and non- HBCU Graduates This study also provides evidence of the net effect of graduating from an HBCU on one s occupational status. For example, in this analysis, graduating from an HBCU was associated with higher levels of occupational status. These results suggest that African American college graduates who have similar educational and personal histories, who are the same with respect to age, and who share similar levels of social and cultural capital are more likely to achieve high status occupations if they graduated from an HBCU. On the one hand, these findings are somewhat consistent with those found in earlier research (Ehrenberg & Rothstein, 1994; London, 1998) and may also reflect prior conclusions that HBCUs tend to foster educational climates that engender African American college student success (Watson et al., 2002). On the other, that African Americans who attend HBCUs achieve higher occupational statuses than those who do not attend such institutions lends support to the continuing significance of HBCUs. Despite the fact that predominately White institutions educate (not necessarily graduate) most Black college students today, HBCUs still award a large majority of all BAs earned by African Americans. This is particularly true in high status career fields such as law, medicine, and science (Brown & Freeman, 2002). Findings from this study may reflect that HBCUs continue to produce the vast majority of black professionals and those whom the black community and society in general have acknowledged as black leaders (Barthelemy, 1984, p. 14).
22 48 The Journal of Higher Education Relationships of Independent Variable with Earnings The evidence also suggests that those who graduated from HBCUs had higher levels of educational aspirations than those who did not attend such institutions. It is also interesting to note that educational aspirations of those who attended HBCUs exceeded the overall B&B sample average including students from other racial/ethnic groups (M = 4.02, SD = 1.07). These estimates are consistent with previous research (Cole, Barber, Bolyard, & Linders, 1999; Pascarella & Terenzini, 2005), and they also provide compelling information about the differences between the educational aspirations of African Americans and students from other groups. However, despite their high aspirations, Black students graduate at lower rates (Nettles & Perna, 1997) and earn less money than their non-black counterparts. That is, despite their high hopes, African Americans tend to earn less. The analysis shows that earnings are a function of age; these findings are consistent with Zhang s conclusion that salary is a concave function of age (2005, p. 322). In this study, controlling for all other factors, predicted earnings are significantly, positively influenced by one s age. For example, a one-unit increase in age is associated with a unit increase in earnings. For example, if an individual who is younger (received BA at 20 years old) earns approximately $15,000 per year, one who is older (received BA at 25 years old) is predicted to earn much more. This may reflect additional compensation for years of experience, but additional investigation is warranted. Previous studies suggest that there is a tipping point in the effect of age on earnings (Cain, Freeman, & Hansen, 1973; Taubman, 1975). Future research might explore this topic more closely and focus on whether the relationship between age and earnings is mediated by time. Family income is associated with higher earnings, although the estimated effect is rather small. A one-unit increase in family income is associated with a increase in annual earnings. Prior research indicates that socioeconomic factors significantly influence educational outcomes and labor market success (Mare, 1980). Results of this study regarding socioeconomic factors generally support prior conclusions and critical views of American education. For example, socioeconomic factors such as family income and parent s level of education continue to influence earnings, enrollment in graduate school, and type of graduate school selected (Zhang, 2005). This proves both theoretically promising and practically problematic as it provides empirical evidence of the (a) importance of sociocultural capital in theory and (b) the continuing disparities between those who have and those who have not. Annual earnings vary across academic majors. For example, predicted early career earnings of graduates majoring in humanities and social