The wage earnings impact of historically black colleges and universities.

• Author: Mykerezi, Elton
• Position: Survey

1. Introduction

   Historically black colleges and universities (HBCUs) were originally founded to provide postsecondary educational opportunities to blacks. Since the end of de jure segregation in public education under the Supreme Court Brown v. The Board of Education of Topeka (1954) ruling, the HBCU mandate has become the subject of considerable debate. The Supreme Court ruling in United States v. Fordice (1992) may be interpreted to indicate that continued state and federal support of HBCUs as racially distinct institutions must be justified by educational benefits beyond the provision of a low-cost educational alternative (Moore 2000). Differential post-college labor market success is an important indicator of the value of unique HBCU benefits and thus, is of considerable interest in terms of postsecondary education policy.

   Previous studies by Ehrenberg and Rothstein (1994) and Constantine (1995, 1998) use data from the National Longitudinal Study of the High School Class of 1972 to examine the impact of HBCU attendance, relative to attendance of other colleges, on the wages of blacks. However, the findings of these studies are somewhat conflicting. Ehrenberg and Rothstein (1994) find a 0-12% wage loss for HBCU attendees, relative to their counterparts who attended other four-year colleges or universities. Constantine (1995), on the other hand, finds that HBCU attendees realize 38% higher wages than black college attendees at other four-year institutions. Constantine (1998) finds that black males who attend HBCUs realize 7% higher wages than black attendees at other four-year colleges and universities, and that wages would be 25% higher for the general college attendee population of black males with HBCU attendance, Black females who attend an HBCU show 15% higher wages relative to black females who attended other four-year colleges, but there are no significant wage gains from HBCU attendance for the general population of black females. The conflicting findings in these studies potentially arise from changes in the differential return on HBCU education with post-college experience in the labor market, as the median age in the Ehrenberg and Rothstein (1994) sample is 25 years, compared to 32 years in the Constantine (1995, 1998) sample.

   This paper augments the existing literature on the wage-earnings impact of blacks' HBCU attendance in two important ways. First, the age-earnings profiles of HBCU attendees and other college-educated blacks are examined using data from a more recent cohort of blacks in the 1979 National Longitudinal Survey of Youth. Specifically, the current study examines the impact of HBCU attendance on the initial post-college wage, on the most recent survey wage observation, and on the average annual growth rate in wages. Measuring labor market success by the age-earnings profiles of individuals over time provides more complete information about patterns of labor market success than does a wage observation at a single point in time. Second, the study explicitly accounts for the impact of location of HBCUs. Of particular note, the distance from one's residence at age 14 years to the nearest HBCU and to other colleges and universities is controlled for in the college selection decision, as are the local economic conditions for the individual's place of residence at age 14.

   Estimates of the differential return to HBCU attendance are developed in the rest of the paper, as follows: Section 2 presents background information on the role and geographic distribution of HBCUs. Section 3 outlines the estimation strategy and empirical specification. Section 4 describes the data and variables used in the analysis. Section 5 presents the empirical results, while section 6 concludes and distills policy implications.

2. The Role and Location of HBCUs

   HBCUs were established to provide equal educational opportunities for black students denied admission to their states' original 1862 land-grant university systems. (1) Currently, 105 HBCUs are nationally or regionally accredited; of these, 89 are four-year colleges--of which 18 were established as 1890 land grants, and 51 are publicly funded.

   HBCUs continue to play a very important role in the education of black Americans by reducing the costs of obtaining a college education and increasing academic success. In 2001, HBCUs matriculated 21.3% of all African-American students enrolled in four-year colleges, awarded 21.5% of all baccalaureate degrees earned by African-Americans nationwide, and awarded master's degrees and first-professional degrees to about 12% of African-Americans who earned graduate degrees (Provasnik and Shafer 2004). In addition, Brazziel (1983) finds that a disproportionate share of black Ph.D. holders attended HBCUs as undergraduates. Blacks at HBCUs are also more likely to persist toward a degree and to report a more rewarding overall college experience relative to their counterparts at mixed institutions (Deskins 1991; Hoffman, Snyder, and Sonnenberg 1992).

   A highly disproportionate share of HBCUs, including the 18 land-grant HBCUs, are located in the South, and many are in or near areas with high concentrations of persistently poor black families (Mykerezi, Mills, and Gomes 2003). HBCUs also appear to have important spillovers onto the economic well-being of their local communities, as decreased physical distance of a community from the nearest HBCU increases the share of adult blacks with college degrees and thereby, increases community-level per capita incomes (Mykerezi and Mills 2005).

3. Empirical Strategy

   The determinants of black male and female earnings have been shown to differ, and, in the current application, the effects of HBCU attendance are estimated separately by gender. As noted, the earnings impacts of HBCU attendance are also estimated at the time of the first post-college job, at the most recent wage observation, and as the average annual growth in wages between the first and most recent wage observation.

   The effect of attending an HBCU is analyzed as a counterfactual comparison. Following Imbens' (2004) terminology, let W be an indicator of college selection, with W = 1 denoting attendance of an HBCU and W = 0 denoting attendance at any other four-year college. [Y.sub.i](1) and [Y.sub.i](0) denote the logarithm of wage (or wage growth) for individual i with and without HBCU attendance, respectively. We are interested in two effects. The average treatment effect (ATE) for all four-year college attendees is defined in Equation 1 as follows:

   [tau] = E[Y(1) - Y(0)]. (1)

   The average treatment effect on the treated (ATT), in this case those attending an HBCU, is defined in Equation 2 as follows:

   [[tau].sub.t] = E[Y(1) - Y(0) | W = 1]. (2)

   For each individual, however, only one of these wages is observable. Unobserved and observed heterogeneity in HBCU choice will potentially bias inferences on wage impacts drawn from the comparison of the two groups. Therefore, estimators that account for unobserved and observed heterogeneity in HBCU choice are employed.

   With observational data there is a concern that institution selection may, in part, depend on unobserved characteristics of individuals that are also correlated with post-college wages. This endogeneity concern is addressed by estimating a standard two-step Heckman (1979) model that takes advantage of the availability of appropriate instruments and assumes joint normality of the error terms. First, a probit model is estimated as follows:

   P(W = 1 | Z) = [PHI](Z[alpha]) + [[epsilon].sub.1], (3)

   where Z is a vector of covariates assumed to be correlated with college choice, [alpha] is a vector of parameters, [[epsilon].sub.1] is a normally distributed random error, and [PHI](Z[alpha]) represents the normal cumulative density function. For samples of HBCU attendees and other four-year college attendees, second-stage equations are then estimated as follows:

   Y = XB + p[??] + [[epsilon].sub.2], (4)

   where X is a vector of observed covariates and p is a parameter. Further, [??] is the estimated inverse Mills ratio, defined as

   [??] = [??](Z[alpha])/1 - [PHI](Z[alpha]),

   where [phi](Z[alpha]) is the normal probability density function. The significance of the parameter associated with the inverse Mills ratio ([lambda]) provides a test for endogeneity of selection of education institution. Failure to reject the null hypothesis that 9 is statistically zero at a conventional level of significance provides evidence that the institution choice is exogenous after controlling for observed covariates. The choice to attend an HBCU then satisfies the conditional exogeneity assumption: (2)

   [Y(1),Y(0)] [perpendicular to] W | X.

   As a result, the treatment effect could be estimated by ordinary least squares (OLS) as

   Y = [tau]W + X[beta] + [[epsilon].sub.3], (5)

   where Y and W are previously defined, X is the vector of observed covariates, [beta] is a vector of parameters, and [[epsilon].sub.3] is a vector of independently identically distributed regression errors. (3) [tau] is an unbiased estimate of the ATE, which is equal to the ATT under the assumption of a constant proportional treatment with selection strictly on observables.

   Under the conditional exogeneity assumption separate ATE and ATT can be estimated without the assumption of joint normality on the error terms of Equations 3 and 4 by utilizing a propensity score. Specifically, first the propensity score is estimated using the probit specification in Equation 3. Then second-stage wage regressions are estimated by weighted least squares (Price 2005). (4)

   In this case, [tau] is an unbiased estimate of the ATE if the weights used are given by Equation 6:

   [omega] = W/[??] + 1 - W/1 - [??], (6)

   and it...