The Review of Black Political Economy

, Volume 40, Issue 4, pp 371–399

What Does a High School Diploma Get You? Employment, Race, and the Transition to Adulthood

Authors

    • Urban Institute, Center on Labor, Human Services, and Population
  • Daniel Kuehn
    • Department of EconomicsAmerican University
Article

DOI: 10.1007/s12114-012-9147-1

Cite this article as:
McDaniel, M. & Kuehn, D. Rev Black Polit Econ (2013) 40: 371. doi:10.1007/s12114-012-9147-1

Abstract

We compare the employment of African American and white youth as they transition to adulthood from age 18 to 22, focusing on high school graduates and high school dropouts who did not attend college. Using OLS and hazard models, we analyze the relative employment rates, and employment consistency, stability, and timing, controlling for a number of factors including family income, academic aptitude, prior work experience, and neighborhood poverty. We find white high school graduates work significantly more than all other youth on most measures; African American high school graduates work as much and sometimes less than white high school dropouts; African American dropouts work significantly less than all other youth. Findings further suggest that the improved labor market participation associated with a high school diploma is higher over time for African Americans than for white youth.

Keywords

High schoolRacial disparitiesEmploymentTransition to adulthood

Introduction

How do employment outcomes differ for youth who receive a high school diploma and those who do not during the transition to adulthood, and does this difference vary by race? We examine how well a high school diploma translates to employment for young adults who do not pursue post-secondary education by comparing employment of African American and white youth who did and did not complete high school.

Roughly thirty percent of the US workforce had a high school diploma, its equivalent, or less in 2010 (BLS 2011b). Among males approaching adulthood (between ages 16 to 24), that share may be closer to half (49.5 %) (Mincy et al. 2006). While the US has witnessed growth in jobs requiring (and workers attaining) post-secondary credentials, a sizeable remaining share of jobs do not require post-secondary credentials (Holzer and Lerman 2007; Holzer 2010; Lacey and Wright 2009). Workers without postsecondary degrees are more vulnerable to unemployment than those with degrees, and since the 1970s a substantial wage gap has emerged between those with a high school diploma and those with a college education (Goldin and Katz 2007). Despite this decline in the value of a high school diploma relative to a college degree, the most vulnerable of all workers remain individuals who have not completed high school. In 2010 the employment and earnings gaps between workers with and without a high school diploma was larger than the employment and earnings gaps separating workers with a high school diploma and an associate’s degree (BLS 2011a). This suggests the credential remains very important. But how important?

We compare employment during young adulthood for high school graduates and dropouts who do not pursue post-secondary education or training. We focus on the initial period of transition from ages 18 to 22—a time when youth who do not attend college are generally securing a foothold in the labor market. Early traction generally leads to later success, and researchers consider this a critical period for establishing and maintaining sustained connection to work (Raaum and Røed 2006).

Employment and earnings in the US have always varied dramatically by race. We hypothesize that the relative value of the high school diploma does as well. Given long standing disparities in employment and earnings between African Americans and whites, the unique history of discrimination against and segregation of African Americans, and the historical use of comparisons between African Americans and whites as a barometer of US inequality, we focus on these two groups using a nationally representative cohort of youth from the National Longitudinal Survey of Youth (NLSY), 1997 who were between the ages of 13 and 16 on December 31st, 1996, and were interviewed annually thereafter.

Attaining wage rates high enough to live independently and attaining consistent employment are essential elements of the transition to adulthood and a notable area of racial disparity. However, we do not focus on earnings and wages in this study. Others have explored these differences, which are largely explained by differences in human capital levels (Neal 2006; Black et al 2006; Neal and Johnson 1996). However, the reasons for disparities in employment rates by race have been more elusive than the reasons for disparities in earnings and wages (Ritter and Taylor 2011), which is why we focus on employment disparities here. Among young African American men two thirds of the decline in work compared to young white men over the last several decades is unexplained, even after adjusting for skills and education (Holzer and Offner 2006). Another reason why we focus on employment rather than earnings is that since our study concerns youth with no higher than secondary education, and the analyses are restricted to their early work careers from ages 18–22, low wage rates may be expected, but early and large differences in employment (particularly full-time employment) may better foreshadow future, long term disparities.

Background

Economic returns on a terminal high school diploma have declined relative to a post-secondary degree in the last six decades as employer demand for and rates of post-secondary education have climbed (Goldin and Katz 2008). The change corresponds with a lengthening of the transition period from adolescence to adulthood and the recognition among psychologists of a developmental stage known as “emerging adulthood” when a youth is no longer considered an adolescent, but is not quite considered an adult (Arnett 2007). This lengthening is an international trend driven by the increased demand for skilled labor and cost of living increases that outpace wage increases (Bell et al. 2007).

When is an individual an adult? Economic self-sufficiency is still a key marker youth identify (Arnett 2001; Mortimer et al 2008), and many postpone its attainment until after they have acquired some post-secondary experience. But delaying economic self-sufficiency until after post-secondary education or training is not an option everyone chooses or is able to afford or attain. Other paths to adulthood include work after leaving or completing high school, and there is some evidence that youth on this path have a somewhat accelerated transition to adult status (Osgood et al 2005; Mortimer et al 2008). Youth who do not attain post-secondary education tend to report sooner that they have begun work that will lead to their eventual career (Mortimer et al 2008).

The path to economic self-sufficiency is not always smooth, particularly for youth right out of high school (Osgood et al 2005). However, having a high school diploma appears to help, and may signal necessary skills to employers (Spence 1973; Belman and Heywood 1997). Having a high school diploma, compared to not having one, is associated with higher lifetime earnings (Day and Newburger 2002) and more steady employment during the transition to adulthood (Finn 2006). Among a national sample of youth at-risk of not completing high school in the US—defined as living in low-SES homes and attending low-SES schools—those with a high school diploma and no post-secondary education worked more consistently (i.e., longer spells of full time employment) eight years after high school than those without a high school diploma, regardless of their grades and standardized test scores in high school (Finn 2006). Within racial groups, youth with a high school diploma or equivalent work more weeks between the ages of 18 and 22 than same-race peers without a high school diploma or equivalent (US Bureau of Labor Statistics 2010).

Among high school graduates, employment varies significantly by race. African Americans with the same level of education fare quite differently in the labor market than whites (McElroy 2005). Nationally, African Americans with a high school diploma or equivalent who have never enrolled in college have worked fewer weeks between the ages of 18 and 22 than white youth with the same credentials (57.8 % weeks worked compared to 76 % weeks worked, respectively) (US Bureau of Labor Statistics 2010). Earnings among workers with a high school diploma are just as disparate by racial group. Among full-time, full-year workers ages 25–34 in 2008, African Americans with a high school diploma had median earnings of $26,000 (in 2008 dollars) compared to $31,200 among whites (Aud et al. 2010).

Common hypotheses for enduring racial differences in employment include differences in values, behaviors, or abilities and racial discrimination within the labor market and within non-labor market activities (e.g., access to job-training programs) (Mason 2005). Among young adults entering the labor market, explanations for lower employment among African Americans compared to whites include lower educational attainment and job skills on average and differences associated with higher concentrated poverty including less job access, different presentation styles, and more criminal involvement (Stoll 2005; Pager et al 2009). Other explanations include poor social networks and job training preparation (Royster 2003; Granovetter 1985).

In-person audit studies, in which two equally-matched job applicants who differ by race apply for the same job, provide support for some that racial discrimination plays a significant role in employment differences. Authors employed this methodology to study the low-wage labor market in New York City and found African Americans were half as likely as equally-matched whites to get a job offer or a callback. African American and Latino applicants with no criminal record fared similarly as whites recently released from prison (Pager et al. 2009). Although scholars debate whether audit studies accurately diagnose discrimination1 (Heckman 1998) or even reflect discrimination experienced by actual job seekers who may systematically locate non-discriminatory employers (Becker 1957), employment disparities have not been explained fully, and racial discrimination remains a plausible explanation.

Conceptual framework and hypotheses

According to the theory of statistical discrimination, employers lacking complete information about a job applicant (e.g., their ability) will attribute to the applicant average perceptions they hold of the applicant’s racial group—often regardless of the applicant’s qualification (e.g. work history, education) (Stoll 2005). Further, when employers do not have complete information on workers, they use group markers (often physical features) to differentiate between them (Phelps 1972; Arrow 1973). Racial disparities can emerge because employers attribute perceived group traits to individuals when an individual’s actual traits and qualities are unknown or unobserved. This implies that if the average African American youth has lower education and less work experience than the average white youth, employers will choose to hire the white applicant, holding all else equal. Based on the theory of statistical discrimination, we hypothesize that among a national sample of youth who do not obtain post-secondary education, African Americans with a high school diploma will have poorer employment outcomes than white youth with a high school diploma.

The theory of statistical discrimination does not immediately suggest an answer to our second question: whether we should expect a high school diploma will confer greater advantage on an African American youth (compared to an African American youth without one) than it will to a white youth compared to a white youth without one. Ritter and Taylor (2011) provide an argument for why employers observing the same educational credentials of African American and white job applicants may continue to rely on racial identity in hiring. They suggest that “majority group managers will be generally less astute in assessing the contributions of minority workers than assessing those of majority group workers (p 38).” This explanation represents a departure from traditional statistical discrimination literature, which assumes employers lack information on all employees equally. Under more traditional assumptions, additional information provided by high school diplomas may reduce racial disparities. But if information asymmetries exist by race, racial disparities may remain, and the signal represented by a high school diploma may mean more to an employer for one group than to another. One hypothesis that emerges from Ritter and Taylor’s (2011) study is that if employers are less able to assess African American job candidates, a high school diploma will be more valuable for those candidates than it will be for white candidates, who are easier for majority group employers to assess. Ritter and Taylor (2011) are careful to emphasize that this effect will dominate for employment differentials, and not for wage differentials (which is typically the subject of research in the returns to education literature). Wages can be adjusted after productivity is revealed on the job, and as such may be less contingent on signals than hiring decisions. Belman and Heywood (1991) explore racial differentials in “sheepskin effects” (the impact of a diploma as a signal, independent of the human capital earned) on earnings. They find that for low productivity signals (e.g., high school diplomas), sheepskin effects on earnings are much weaker than for high productivity signals (e.g., college degrees). These findings are consistent with Ritter and Taylor’s (2011) distinction between an educational signal’s effect on earnings versus employment.

We examine whether employment outcomes associated with a high school diploma differ for African American and white youth. We reference signaling theory but do not test it directly since other contributors including differences in human capital (e.g., quality and length of schooling) are also relevant and handled with less precision in our models. If information asymmetry is present for all workers (as in Phelps 1972 and Arrow 1973), but higher for African American youth (as in Ritter and Taylor 2011), we would expect African American youth in general to have lower employment rates, but a greater differences in employment between youth with and without a high school diploma than white youth with and without a high school diploma.

We anticipate significant employment differences by race even after adjusting for confounding factors associated with employment outcomes such as household income, family structure, parent’s employment and education, and youth characteristics including academic achievement, mental health, social behaviors, incarceration, and work history (Gresenz and Sturm 2004; Holzer et al. 2005; Holzer 2009). Holding these and other factors constant, we examine the relationship between race, high school completion, and employment outcomes during the transition to adulthood.

Data and methods

We use data from the National Longitudinal Survey of Youth 1997 cohort (NLSY97)—a nationally representative sample of approximately 9,000 youths who were 12 to 16 years old on December 31, 1996 and interviewed annually. The NLSY97 over-samples and re-weights African American and low-income youth for more precise estimates of these groups. The sample for this study is restricted to African American and white youth who were 13 to 16 years old on December 31st, 1996, and who did not attend college (2-year or 4-year) at any time before their last round of interviews. We excluded youth who were 12 during their first interview, some of whom had not turned age 22 by their 2005 interview. Youth included in the analyses also did not have missing information on the dependent employment variables of interest.

Nearly two out of five (39 %) African American and white youth who were ages 13–16 in 1996 had not enrolled in a two- or four-year college by their 2005 interview. The final analysis sample includes 612 African American and 891 white youth observed through 2005 when the youngest had turned age 22. We compare employment patterns for African Americans and whites, and for high school graduates and dropouts.

Dependent variables

We focus on relative employment rates, and employment consistency, stability, and timing during the transition to adulthood. We examine the proportion of weeks a youth was employed either part time or full time in the four-year period between ages 18 and 22, and the proportion of weeks he or she was employed full time at age 18 and at age 22. For employment timing we estimate how long youth take to attain steady employment, or six consecutive months of full time work after their 18th birthday.2

All employment measures are derived from detailed job history variables that provide information on a youth’s employment status and hours worked every week since his or her first interview in 1997. A youth who works 35 h per week is considered a full time worker although the hours may come from more than one job; part time is considered anything less than 35 h per week. The NLSY requests information on both formal and informal work, so informal employment that would be missed in administrative data is recovered in these data. It is possible that informal work could be somewhat underreported.

Independent variables

Our primary variables of interest are race and high school completion. Youth are considered African American if that is the single race they used to identify themselves and reported being non-Hispanic. Youth are considered white if that is the single race they used to identify themselves and reported being non-Hispanic. Youth who identified themselves as multi-racial or Hispanic or any race other than African American or white were not included in the analyses. A youth is considered a high school graduate if he or she reported earning a high school diploma at any time before the last round of interviews. Youth who do not complete high school and report earning a general equivalence degree (GED) are included in the dropout category, to be consistent with findings from Cameron and Heckman (1993) and Heckman et al. (2010) that GEDs are not equivalent to a high school diploma in the eyes of employers. We compared employment outcomes among high school dropouts with and without a GED to confirm this research, and generally found very little differences. With respect to the total time spent working between ages 18 and 22 (i.e., percent of weeks worked between ages 18 and 22), high school dropouts with a GED look more similar to high school dropouts without a GED than to high school graduates. Another reason for including GED holders with dropouts is that the GED can be considered an outcome of dropping out, along with employment, exiting the labor market, or the pursuit of job training. This assumption that GED holders are most comparable to high school dropouts is relaxed in the analyses presented in Appendix 1.

Additional covariates

Our models account for several factors beyond race and high school completion that could contribute to youth employment. One variable of interest is the youth’s family income during the first wave of the survey. Family income is expressed as a ratio to the federal poverty level. We use parents’ earnings and other income, household size, and the 1996 poverty thresholds to create income-to-poverty ratios for each family. Our measure is specifically a parental income measure, since we omitted earnings from other members of the household. Two types of missing data required adjustments to our parental income variable. In the survey, one parent responds for the family. In some cases, the responding parent does not report the other parent’s earnings or gives an incomplete account of “other income”. In addition, about 10 % of youth do not have any parent report, and therefore all family income information is missing. To handle these missing data issues, we used 1997 earnings and other family income reported by the parents in 1998 (deflated to 1996 dollars using the consumer price index CPI). We replaced cases with any missing 1996 income data with 1997 income data if it was complete. If income data was not complete for either 1996 or 1997 it remained missing. This approach eliminated most of the missing values problem. We also included a dummy variable for youth with no family income reported to control for any differences in these cases. Another important limitation of the income variable is that family income is measured for only one year. This restricted observation period prevents us from taking into account the significant transitory nature of income, particularly among low-income households who may move in and out of poverty.

A second independent variable was the youth’s family structure. A categorical indicator was created for youth who grew up in a single parent family, a family with two parents but only one biological, families with two biological parents, and all other family structures.

Incarcerated youth are much less likely to be working, so we included a variable that was equal to one if a youth spent any time in jail between the ages of 18 and 19, and zero if they did not. We also assessed risky behavior during adolescence with a cumulative risk score incorporating thirteen possible risk behaviors (possible values for this cumulative risk behavior score range from zero to 13). The risk behaviors covered by our cumulative score are: consumed alcohol by age 13, used marijuana by age 16, ever used other drugs, engaged in sex by age 16, ever attacked someone and/or got into a fight, ever been a member of a gang, ever sold drugs, ever destroyed property, ever stole something worth less than $50, ever stole something worth more than $50, ever committed another type of property crime (i.e. vandalism), ever carried a gun, and ever ran away from home. To check the reliability of summing all thirteen behaviors for this measure, we conducted a confirmatory factor analysis. The Cronbach alpha between the factor score and the cumulative risk measure was very high (0.98), suggesting that the cumulative measure of risk does indeed hold together as a singular risk measure.

Analyses also control for individual characteristics (gender, academic aptitude, mental health, percent of weeks employed between ages 16 and 18, and incarceration), and family characteristics (parent education and employment, family structure, household size, and receipt of cash welfare in the five years prior to the first round). We also control for whether females have given birth, since their employment outcomes may differ both from females who have not given birth and from males. We also adjust for region of the country (east, north central, south, and west) and whether youth lived in a rural area (compared to non-rural). Youth in the sample turned 18 over a four year period from 1998 to 2001. Those turning 18 earliest experienced more years of a healthy economy before the recession of 2001, so we account for the year youth turned 18 in our models. It is worth noting that the analysis follows a single cohort at a specific point in time, which may not be generalizable to other time periods with relatively stronger or weaker economies. In fact, gaps between African Americans and whites appear to widen during a recession (Couch and Fairlie 2010; Holzer and Hlavac 2012). This suggests that youth who turned 18 after our current cohort from the NLSY—particularly those who turned 18 during the great recession of 2007–2009—could potentially have larger employment disparities than our findings indicate. Research on the returns to schooling confirms that academic test scores are an important and appropriate proxy for ability and should be used in models examining the effects of schooling (McKinley et al. 1995). We therefore included a measure of academic ability based on scores on the Armed Services Vocational Aptitude Battery (ASVAB), which is measured in the baseline year of the NLSY97. Respondents were asked to take a standardized test used by the military for determining enlistment acceptability, consisting of ten subtests. Four of these subtests measure verbal and math ability and when combined, provide a measure (the Armed Forces Qualifying Test, or AFQT score) that correlates highly with standard IQ tests. These scores are available for approximately 79 % of the youth. We included a dummy variable for respondents who chose not to take the ASVAB.

Mental health problems are measured using the Mental Health Inventory-5 (MHI-5). The MHI-5, administered to NLSY97 respondents in 2000, 2002 and 2004, is a set of five questions used to assess degrees of depression and anxiety. The MHI-5 has been used in a number of studies and has been shown to be a valid measure of depression and anxiety among adolescents and adults (Ostroff et al. 1996; Berwick et al. 1991). To assess mental health as close to adolescence as possible, we used the mental health score from 2000. If the mental health score was missing in 2000, the score from the 2002 survey was used.

Most covariates included in the analyses were available in the public-use NLSY97 data file, however neighborhood characteristics (e.g., census tract poverty rates) were not. We accessed restricted-use local level geography (e.g., zip codes and census tracts) data on the youth from the Bureau of Labor Statistics in order to create a “distressed neighborhood” variable that equals one for neighborhoods with a poverty rate greater than 30 %, and zero for neighborhoods with a poverty rate below 30 %.

Methods

We use both bivariate and multivariate methods to examine our primary research questions. We conduct comparison of means tests (t-tests) to check for racial differences in the demographic, family, and youth characteristic measures used. We also estimate the proportion of weeks youth were employed between their 18th and 22nd birthdays. To examine differences in employment timing we create a cumulative survival graph to map the percent of youth at each age from 18 through 22 who have begun a spell of full-time employment that will last at least 6 consecutive months. These bivariate comparisons show whether employment outcomes vary by race and high school completion, and provide a first look at whether and how disparate youth are in employment as they transition to adulthood.

We use ordinary least squares (OLS) regression and parametric hazard models to examine the independent contribution of race and a high school diploma after taking demographic, family, and youth characteristics into account. Models assessing the proportion of weeks worked over a period of time are estimated using OLS regression.3 Models assessing the time until youth attain steady full time employment are estimated using accelerated failure time models—a class of parametric hazard models. We use the NLSY97 longitudinal design and continuously-timed employment information to examine the elapsed time until youth begin a spell of steady full-time employment after their 18th birthday. Youth are observed from the week they turn 18 until the week they begin the steady employment spell. A censored youth is one who does not begin steady work within our observation period. In 2005 when youth in our sample were last observed, they ranged in age from 22 to 25. Therefore, some youth were observed through age 22 while others were observed through age 25. Hazard models are ideal because they take into account potential bias from censored data and varying observation periods (Allison 2008). We calculate the employment “time ratios” based on a Weibull distribution.4

For both the OLS and hazard regression analyses we run four sets of models adding progressively more covariates. The first models examine the relationship between race and high school graduation status on the outcomes of interest without other covariates. We add demographic variables including youth’s age, gender, family poverty, region, and rural residence in the second set of models. We add parent and family characteristics in the third set of models including parent’s education, employment, family structure, family receipt of cash assistance, and residence in a distressed neighborhood. In the fourth set of models we add youth characteristics including academic aptitude score (on the AFQT military entrance exam), mental health, female adolescent childbearing, employment between ages 16 and 18, incarceration, and cumulative risk behavior score. The full models in each set of analyses follow the equations below:
$$ \mathrm{OLS}\;\mathrm{Regression}:\mathrm{Emp}=\upalpha +{\upbeta_1}\mathrm{Race} {-}\mathrm{Gra}{{\mathrm{d}}_{\mathrm{i}}}+{\upbeta_2}\mathrm{Dem}{{\mathrm{o}}_{\mathrm{i}}}+{\upbeta_3}\mathrm{Fa}{{\mathrm{m}}_{\mathrm{i}}}+{\upbeta_4}\mathrm{Cha}{{\mathrm{r}}_{\mathrm{i}}}+{\upbeta_5}\mathrm{Nhoo}{{\mathrm{d}}_{\mathrm{i}}}+{\varepsilon_{\mathrm{i}}} $$
(1)
$$ \mathrm{Parametric}\;\mathrm{Hazard}:\mathrm{Log} \ \mathrm{T}=\upalpha +{\upbeta_1}\mathrm{Race}{-}\mathrm{Gra}{{\mathrm{d}}_{\mathrm{i}}}+{\upbeta_2}\mathrm{Dem}{{\mathrm{o}}_{\mathrm{i}}}+{\upbeta_3}\mathrm{Fa}{{\mathrm{m}}_{\mathrm{i}}}+{\upbeta_4}\mathrm{Cha}{{\mathrm{r}}_{\mathrm{i}}}+{\upbeta_5}\mathrm{Nhoo}{{\mathrm{d}}_{\mathrm{i}}}+{\varepsilon_{\mathrm{i}}} $$
(2)

Where dependent variable, Emp, is the percent weeks youth i was employed at age 18 or at age 22; and dependent variable, Log T, is the log time until youth i attains steady full time employment. Race-Gradi is a vector of variables indicating the youth’s race and graduation status, including white dropouts, African American graduates, and African American dropouts (white graduates are the reference category). Demoi is a vector of the youth’s demographic characteristics, Fami are the youth’s parent and family characteristics, and Chari are the youth’s personal characteristics. A random disturbance term in each equation is represented by εI and a constant term by α.

Results

Table 1 shows selected sample characteristics for African American and white high school graduates and dropouts prior to age 18. The proportion of youth in the sample (i.e., those with a high school diploma, its equivalent, or less) who did not complete high school varied by race. Among African Americans, 25 % did not complete high school compared to 15 % of whites. The proportion of high school graduates who never enrolled in college also varied by race—with African American youth significantly less likely to enroll (38 % never enrolling compared to 27 %).
Table 1

Sample characteristics by race and education

 

White HS Graduates

African American HS Graduates

White HS Dropouts

African American HS Dropouts

n = 2034

n = 893

n = 370

n = 307

No college enrollment (2- or 4-year)abcde

0.27

0.38

0.91

0.91

Study Sample (no college enrollment)

n= 554

n= 332

n= 337

n= 280

% of weeks employed full time, age 18abcdf

0.57

0.4

0.44

0.27

% of weeks employed full time, age 22abcdf

0.64

0.57

0.54

0.37

General Demographics

 Age in 1997c

14.6

14.5

14.4

14.5

 % Age 18 in 2001

0.24

0.26

0.28

0.26

 % Age 18 in 2000

0.27

0.28

0.29

0.25

 % Age 18 in 1999f

0.21

0.21

0.19

0.26

 % Age 18 in 1998

0.28

0.25

0.24

0.23

 Female

0.43

0.45

0.45

0.39

 % Federal Poverty Levelabcdf

2.62

1.68

1.91

1.1

 Northeastab

0.12

0.14

0.18

0.20

 North Centralabef

0.34

0.16

0.31

0.18

 Southabef

0.29

0.62

0.34

0.65

 Westabef

0.17

0.09

0.17

0.05

 Ruralabcef

0.38

0.23

0.31

0.18

Parent Characteristics

 Parents less than HSbcde

0.11

0.11

0.29

0.29

 Parents HS onlycef

0.59

0.62

0.48

0.61

 Parent some collegebcdf

0.30

0.27

0.22

0.10

 Parent any FT employmentabcef

0.87

0.69

0.80

0.62

 Two parent, both biologicalabcf

0.53

0.25

0.29

0.18

 Two parent, one biologicalcef

0.19

0.14

0.23

0.14

 Single parentabcef

0.25

0.51

0.41

0.58

 Household sizebdf

4.3

4.4

4.3

4.9

 Parent received AFDCabcdf

0.22

0.43

0.43

0.55

 Distressed neighborhoodabdef

0.02

0.27

0.02

0.35

Youth Characteristics

 AFQT (military entrance score)abcdef

36.2

20.8

23.8

10.4

 AFQT missingabcdef

0.15

0.18

0.25

0.23

 Mental health scorebcde

14.9

14.8

14.3

14.3

 Teen child birthbcd

0.01

0.03

0.05

0.08

 % Employed age 16–18abef

0.54

0.39

0.48

0.31

 Any jail age 18–19bcde

0.02

0.02

0.10

0.10

 Cumulative risk scoreabcdef

4.0

3.2

5.6

4.5

ablack grad significantly different than white grad (p < .05)

bblack dropout significantly different than white grad (p < .05)

cwhite dropout significantly different than white grad (p < .05)

dblack grad significantly different than black dropout (p < .05)

eblack grad significantly different white dropout(p < .05)

fblack dropout significantly different than white dropout (p < .05)

Compared to white high school graduates, African American graduates tended to grow up in poorer families, were much more likely to live in the South and in non-rural and distressed communities. A lower proportion of African American graduates compared to white high school graduates had a parent who was employed full time (69 % compared to 87 %), and more African American graduates lived with a single parent (51 % compared to 25 %). African American high school graduates also had lower academic aptitude scores based on the AFQT military entrance exam, and worked less prior to age 18 (39 % ever employed between ages 16–18 compared to 54 % of white high school graduates). It is important to note that fewer African American than white youth took the aptitude test, and fewer dropouts than graduates took the test. With respect to risky behaviors, African American high school graduates reported engaging in fewer on average than white high school graduates (3.2 compared to 4.0).

African American high school dropouts appear most disadvantaged prior to age 18 in comparison to the other three groups of youth—they grew up in significantly poorer households, were most likely to live in single-parent families in larger households, and to live in distressed communities. However, they are more similar to African American high school graduates than to white youth with respect to employment between ages 16–18 and parents’ employment. For example, 31 % of African American dropouts worked at some time between ages 16 and 18 compared to 39 % of African American high school graduates (this difference is not statistically significant).

Figure 1 shows the proportion of African American and white high school graduates and dropouts who have worked full time for six consecutive months at each age between 18 and 22. African American youth take longer than white youth on average to begin a period of full time work lasting 6 months or more. Approximately 60 % of white high school graduates have begun a 6 month spell of full time employment before they turn age 19 compared to approximately 40 % of African American high school graduates who have done so by 19, which is itself a lower rate than the approximately 45 % of white high school dropouts who have done so by age 19. Compared to white high school graduates, white dropouts take over 6 months longer, African American high school graduates take a year longer (age 20) and African American dropouts a year and a half longer (age 20½ for 60 % of these groups to have achieved this indicator of steady employment). At age 22, African American high school graduates appear to catch up to white high school dropouts, and have narrowed the gap, but not caught up to, white high school graduates in the rate of full time employment (from around 60 % vs. around 40 % at age 19, to around 85 % vs. less than 80 % at age 22). In contrast, African American dropouts consistently take longer than African American and white high school graduates and white dropouts to attain steady full time employment. That gap does not change appreciably over this time relative to the white graduates or dropouts, and widens relative to African American high school graduates.
https://static-content.springer.com/image/art%3A10.1007%2Fs12114-012-9147-1/MediaObjects/12114_2012_9147_Fig1_HTML.gif
Fig. 1

Hazard rate of first spell of steady full time work after age 18

We hypothesize differences such as family income, parents’ employment and education, neighborhood quality and safety, academic test scores, and prior work history may explain the discrepancies shown in the descriptive analyses above (Table 1 and Fig. 1). To test the independent contribution of race and a high school diploma on employment consistency, stability, and timing we adjust for the demographic, parent- and youth- related differences in OLS regression and parametric hazard models below.

Proportion of weeks employed full time at age 18

We present four models in Table 2 explaining differences in the proportion of weeks youth work full time at age 18. Model 1 excludes covariates and shows African American high school graduates, African American dropouts, and white dropouts work significantly fewer weeks than white high school graduates. African American graduates, white dropouts, and African American dropouts work 18, 13, and 30 percentage points fewer weeks of full-time employment than white graduates. Model 2 adjusts for demographic factors including household poverty ratio, age, gender, region, and whether a youth lives in a rural area. Although income and gender help account for some of the difference in weeks worked, race and high school completion remain statistically significant with African American graduates, African American dropouts, and white dropouts working significantly fewer weeks in the four-year period than white high school graduates. Gender does not vary substantially across race and education groups (see descriptive statistics in Table 1), suggesting that while gender may be an important determinant of employment, controlling for gender does not account for the difference between graduates and dropouts. White high school graduates come from much higher income families than other youth in the sample (although they are still relatively low income), so the modest reduction in employment disparities at age 18 by race is most likely attributable to controlling for income.
Table 2

OLS regression models explaining total percent of weeks employed full time (age 18) by race and education

 

MODEL 1

MODEL 2

MODEL 3

MODEL 4

Coefficient

SE

Coefficient

SE

Coefficient

SE

Coefficient

SE

African American HS Graduate

−0.176***

0.028

−0.140***

0.030

−0.127***

0.033

−0.082**

0.033

African American HS Dropout

−0.301***

0.028

−0.263***

0.032

−0.242***

0.035

−0.168***

0.036

White HS Dropout

−0.130***

0.027

−0.105***

0.027

−0.089***

0.028

−0.056**

0.028

White HS Graduate (Reference)

% Federal Poverty Level

  

0.026***

0.008

0.020**

0.009

0.019**

0.009

% Federal Poverty Level (Squared)

  

0.000*

0.000

−0.000

0.000

−0.000

0.000

% Federal Poverty Level (Missing)

  

0.022

0.040

0.025

0.041

0.017

0.039

Age 18 in 2000

  

0.033

0.029

0.034

0.029

0.027

0.003

Age 18 in 1999

  

0.065**

0.031

0.061**

0.031

0.071**

0.030

Age 18 in 1998

  

0.013

0.030

0.017

0.030

0.068**

0.311

Age 18 in 2001 (Reference)

Female

  

−0.135***

0.022

−0.129***

0.022

−0.124***

0.023

North Central

  

0.002

0.033

0.006

0.033

0.004

0.032

South

  

−0.028

0.031

−0.025

0.031

−0.016

0.030

West

  

−0.059

0.039

−0.056

0.040

−0.046

0.039

Northeast (Reference)

Rural

  

0.044**

0.024

0.040*

0.024

0.040*

0.024

Parents less than HS

    

−0.051

0.035

−0.018

0.035

Parents HS only

    

−0.011

0.027

0.014

0.027

Parents some college or higher (Reference)

Parent any FT employment

    

0.034

0.028

0.030

0.028

Single parent

    

−0.018

0.032

−0.015

0.031

Two parent, one biological

    

−0.003

0.029

−0.004

0.028

Other family structure

    

−0.041

0.050

−0.092

0.048

Two biological parents (Reference)

Household size

    

0.016**

0.008

0.015**

0.007

Parent received AFDC

    

−0.031

0.025

−0.024

0.024

Distressed neighborhood

    

−0.017

0.035

0.003

0.034

Distressed neighborhood (Missing)

    

−0.049

0.134

−0.051

0.102

AFQT (Military Entrance Score)

      

0.001

0.001

AFQT (Missing)

      

0.019

0.031

Mental health score

      

0.017***

0.005

Teen child birth (Female only)

      

0.050

0.058

% Employed age 16 − 18

      

0.241***

0.036

Any jail age 18 − 19

      

−0.171***

0.039

Cumulative risk score

      

0.005

0.003

Constant

0.572***

0.017

0.546***

0.043

0.479***

0.067

0.011

0.106

 Sample Size

1469

 

1469

 

1469

 

1469

 

 Adjusted R-Squared

0.062

 

0.112

 

0.121

 

0.1785

 

* p < .10; **p < .05; ***p < .01

In Model 3 we add household family factors including parents’ education, parents’ employment, family structure and size, welfare receipt, and whether the family lived in a distressed neighborhood at the first interview. Growing up in a single parent household and having a parent who did not graduate from high school was weakly associated with fewer weeks worked, however, those factors did not explain the continued differences by race and education. In Model 4 we add youth characteristics including academic aptitude, mental health, female adolescent childbearing, prior employment before age 18, any incarceration between ages 18–19, and engagement in adolescent risk behaviors. These additional factors likewise did not completely explain the variation in employment among African American and white high school graduates and dropouts. In the full Model 4 African American high school graduates work 8 % fewer weeks than white graduates, while white dropouts work 6 % fewer, and African American dropouts work 17 % fewer weeks than white graduates.

Several of the youth characteristics added in Model 4, such as incarceration or risky behavior, introduce an important endogeneity problem. While each plausibly contributes to lower employment rates, poor employment prospects or expectations of poor prospects are often a key motivation for engaging in risky behavior or criminal activity (Borjas et al. 2010). These coefficients are therefore not clear evidence of a causal relationship. Engagement in risky behavior is also self-reported, and there is some mixed evidence that African Americans may underreport criminal activity (Farrington et al. 1996; Thornberry and Krohn 2003). The most striking difference between African Americans and whites among the characteristics added in Model 4 was the difference in their employment rates between ages 16 and 18 (differences in other variables added in Model 4 exhibited greater variation across education groups, rather than race groups). Controlling for teenage employment is therefore likely to be the most substantial reason for the decline in the returns to a high school diploma, by race.

Time until steady full time employment during the transition to adulthood

We present four models in Table 3 showing the time it takes African American high school graduates, African American high school dropouts, and white high school dropouts to attain six consecutive months of full time employment after age 18 compared to white high school graduates. Model 1 (column 2) excludes covariates and shows the relative time difference by race and graduation status without adjusting for additional factors. The results show African American high school graduates take twice as long as white high school graduates to attain steady full time employment. This is longer than it takes white high school dropouts, who take 70 % longer than white high school graduates. African American dropouts take 4.1 times longer than white high school graduates to attain 6 months of steady full time employment.
Table 3

Parametric hazard models explaining time until full time steady employment by race and education

 

MODEL 1

MODEL 2

MODEL 3

MODEL 4

TM Ratio

SE

TM Ratio

SE

TM Ratio

SE

TM Ratio

SE

African American HS Graduate

2.03***

0.322

1.91***

0.338

1.83***

0.343

1.49*

0.281

African American HS Dropout

4.11***

0.731

3.68***

0.716

3.56***

0.768

2.42***

0.537

White HS Dropout

1.70***

0.265

1.48**

0.229

1.36*

0.217

1.12

0.182

White HS Graduate (Reference)

% Federal Poverty Level

  

0.88***

0.042

0.87**

0.048

0.87**

0.048

% Federal Poverty Level (Squared)

  

1.00

0.001

1.00

0.001

1.00

0.001

% Federal Poverty Level (Missing)

  

0.90

0.249

0.85

0.235

0.92

0.260

Age 18 in 2000

  

0.87

0.148

0.88

0.152

0.93

0.154

Age 18 in 1999

  

0.89

0.172

0.93

0.179

0.89

0.169

Age 18 in 1998

  

1.12

0.201

1.11

0.200

0.92

0.172

Age 18 in 2001 (Reference)

Female

  

1.95***

0.254

1.95***

0.256

1.88***

0.260

North Central

  

1.16

0.225

1.14

0.221

1.23

0.235

South

  

1.11

0.199

1.08

0.195

1.08

0.193

West

  

1.57*

0.357

1.58**

0.365

1.67**

0.392

Northeast (Reference)

Rural

  

0.83

0.119

0.83

0.121

0.82

0.120

Parents less than HS

    

1.29

0.273

1.19

0.260

Parents HS only

    

0.99

0.157

0.92

0.149

Parent some college or higher (Reference)

Parent any FT employment

    

0.93

0.666

0.91

0.170

Single parent

    

1.03

0.188

1.00

0.180

Two parent, one biological

    

0.89

0.152

0.86

0.144

Other family structure

    

1.48

0.438

1.18

0.326

Two biological parents (Reference)

Household size

    

0.93*

0.042

0.93*

0.042

Parent received AFDC

    

1.07

0.156

1.07

0.155

Distressed neighborhood

    

1.03

0.200

0.92

0.173

Distressed neighborhood (Missing)

    

1.51

1.866

1.44

1.593

AFQT (Military Entrance Score)

      

1.00

0.003

AFQT (Missing)

      

0.73*

0.138

Mental health score

      

0.92***

0.026

Teen child birth (Female only)

      

1.23

0.512

% Employed age 16–18

      

0.25***

0.052

Any jail age 18–19

      

1.82*

0.435

Cumulative risk score

      

0.99

0.021

 Sample Size

1503

 

1503

 

1503

 

1503

 

 /ln_p

−0.68***

0.023

−0.66***

0.023

−0.66**

0.023

−0.64***

0.023

 p

0.50

0.012

0.51

0.012

0.52

0.012

0.53

0.012

 1/p

1.98

0.045

1.94

0.044

1.94

0.044

1.90

0.044

* p < .10; **p < .05; ***p < .01

When demographic factors are added in Model 2 (column 3), higher family income is associated with a shorter time to attaining steady full time employment and being female is associated with a 96 % increased time. As with differences in employment at age 18, the reduction in racial differences after demographic factors are added are primarily attributable to family income, rather than gender, since family income varies the most substantially across race and education groups. Nevertheless, differences by race and high school completion remain after adding these demographic factors. African American high school graduates take 91 % longer, white dropouts take 48 % longer, and African American dropouts take 3.7 times longer than white high school graduates after taking family income, gender, and region into account. In Model 3 (column 4) we add the set of parent- and family- factors described above. Alone, none of these factors has a significant effect on employment timing (at p < .05), and when included contributes to only a modest reduction in the effect of race and a high school diploma.

In model 4 (column 5) we add youth characteristics to the model. Youth with a higher mental health score and youth who have been employed between ages 16 to 18 attain 6 months of steady full time employment in less time on average than youth with lower mental health scores or who have not been employed between ages 16 to 18. On the other hand, youth who have been incarcerated (a characteristic that varies dramatically between educational groups, although not racial groups) between ages 18 and 19 take 82 % longer than youth who have not been incarcerated to attain 6 months of steady full time employment. When we include these youth characteristics in the model we see additional decline in the race and diploma effect, with most of the decline attributable to teenage employment (which varies the most substantially across race groups). However, African American high school graduates and dropouts still take longer (49 % and 242 %, respectively) than white high school graduates to attain 6 months of steady full time employment. The difference between white high school graduates and white dropouts is not statistically significant in model 4 once youth characteristics are taken into account.

Proportion of weeks employed full time at age 22

White high school graduates work significantly more full time weeks at age 18 compared to African American graduates, African American dropouts, and white dropouts (Table 2), and the differences are not explained by a host of demographic, family, and youth characteristics including academic aptitude and prior work experience. They also secure steady full time employment earlier than African American youth (graduates and dropouts), and those differences are not explained by demographic, family, and youth characteristics (Table 3). A remaining question is whether African American graduates, African American dropouts, and white dropouts eventually work as many weeks as white high school graduates by the time they are 22.

In Table 4 we present findings from OLS regression models explaining the proportion of weeks youth work full time at age 22 (i.e., until 23rd birthday). Without adjusting for covariates (model 1) African American graduates, African American dropouts, and white dropouts work significantly fewer weeks than white graduates. African American graduates work 8 % fewer weeks than white high school graduates, white dropouts work 10 % fewer weeks, and African American dropouts work 27 % fewer weeks.
Table 4

OLS regression models explaining percent of weeks employed full time at age 22 by race and education

 

MODEL 1

MODEL 2

MODEL 3

MODEL 4

Coefficient

SE

Coefficient

SE

Coefficient

SE

Coefficient

SE

African American HS Graduate

−0.079**

0.033

−0.05

0.034

−0.05

0.038

−0.024

0.038

African American HS Dropout

−0.273***

0.034

−0.245***

0.038

−0.236***

0.042

−0.185***

0.044

White HS Dropout

−0.103***

0.019

−0.083***

0.029

−0.068**

0.030

−0.05

0.030

White HS Graduate (Reference)

% Federal Poverty Level

  

0.030***

0.009

0.027***

0.010

0.027***

0.010

% Federal Poverty Level (Squared)

  

−0.001***

0.000

−0.001**

0.000

−0.001**

0.000

% Federal Poverty Level (Missing)

  

0.04

0.044

0.04

0.045

0.030

0.045

Age 18 in 2000

  

0.066**

0.033

0.068**

0.033

0.064**

0.032

Age 18 in 1999

  

0

0.035

0

0.035

0

0.035

Age 18 in 1998

  

−0.01

0.035

−0.01

0.035

0.03

0.036

Age 18 in 2001 (Reference)

Female

  

−0.226***

0.024

−0.221***

0.024

−0.213***

0.026

North Central

  

−0.057*

0.035

−0.06

0.035

−0.062*

0.034

South

  

−0.003

0.033

0.000

0.033

0

0.033

West

  

−0.01

0.043

−0.02

0.043

−0.01

0.043

Northeast (Reference)

        

Rural

  

0.054*

0.026

0.051*

0.026

0.048*

0.026

Parents less than HS

    

−0.05

0.039

−0.03

0.039

Parents HS only

    

0.02

0.029

0.03

0.029

Parent some college or higher (Reference)

Parent any FT employment

    

0.02

0.031

0.01

0.031

Single parent

    

−0.06

0.035

−0.05

0.034

Two parent, one biological

    

−0.04

0.032

−0.03

0.031

Other family structure

    

−0.07

0.052

−0.05

0.051

Two biological parents (Reference)

Household size

    

0

0.008

0

0.008

Parent received AFDC

    

0.020

0.027

0.03

0.026

Distressed neighborhood

    

0.019

0.041

0.03

0.040

Distressed neighborhood (Missing)

    

0.094*

0.056

0.05

0.055

AFQT (Military Entrance Score)

      

0.000

0.001

AFQT (Missing)

      

0.03

0.035

Mental health score

      

0.019***

0.005

Teen child birth (Female only)

      

–0.02

0.063

% Employed age 16–18

      

0.187***

0.038

Any jail age 18–19

      

−0.06

0.049

Cumulative risk score

      

0

0.004

Constant

0.644

0.019

0.662***

0.05

0.671***

0.072

0.260**

0.113

 Sample Size

1418

 

1418

 

1418

 

1418

 

 R-Squared

0.04

 

0.13

 

0.13

 

0.16

 

* p < .10; **p < .05; ***p < .01

Demographic factors are taken into account in model 2 and the differences between African American graduates and white graduates are no longer statistically significant. The models indicate that household poverty may better explain the initial differences between African American and white high school graduates. These factors reduce, but do not explain the differences between African American dropouts, white dropouts, and white graduates however. African American dropouts work 25 % fewer weeks, and white dropouts work eight percent fewer weeks than white graduates. The differences are further reduced but not eliminated with family factors added to the model (model 3). Unlike demographic factors, the family-related factors do not contribute significantly to the models. When youth characteristics are included in the models (model 4), the differences between white graduates and dropouts are no longer statistically significant. However, African American dropouts continue to have significantly fewer weeks of full time employment at age 22 compared to white high school graduates. The covariates that contribute the most to reducing differences in employment outcomes by race are, as with the models presented in Table 2, family income and teenage employment experiences. These characteristics vary the most substantially across racial groups and are significantly predictors of employment at age 22 in the models.

The significance of a high school diploma by race

At age 22 family and youth characteristics better explain the differences in employment among African American high school graduates, white high school graduates, and white dropouts than race or the high school diploma. This is not true for African American dropouts, who continue to work full time significantly less than all other groups despite holding family and youth characteristics constant (Table 4). The findings suggest a high school diploma may confer differential benefits by race—i.e., the difference in employment between African American high school graduates and dropouts may be significantly greater over time than the difference among white high school graduates and dropouts. We examine this and present the results in Fig. 2, which shows the percentage point difference in weeks worked full time by high school graduates compared to same-race high school dropouts at each age. The percentage point estimates come from the full models presented in Tables 3 and 4 (at ages 18 and 22) and similar analyses at ages 19, 20, and 21 (not shown). The comparisons between African American graduates and dropouts come from the same models (Tables 3 and 4 and at each age) with African American graduates as the reference category (not shown).
https://static-content.springer.com/image/art%3A10.1007%2Fs12114-012-9147-1/MediaObjects/12114_2012_9147_Fig2_HTML.gif
Fig. 2

At age 18: African American graduates—dropouts statistically different at p < .01; white graduates—dropouts statistically different at p < .06; the difference between African American graduates and African American dropouts is not statistically different than the difference between white graduates and white dropouts. At age 19: African American graduates—dropouts statistically different at p < .05; white graduates—dropouts statistically different at p < .001; the difference between African American graduates and African American dropouts is not statistically different than the difference between white graduates and white dropouts. At age 20: African American graduates—dropouts statistically different at p < .001; white graduates—dropouts statistically different at p < .001; the difference between African American graduates and African American dropouts is not statistically different than the difference between white graduates and white dropouts. At age 21: African American graduates—dropouts statistically different at p < .001; white graduates—dropouts statistically different at p < .05; the difference between African American graduates and African American dropouts is not statistically different (i.e., not significant at p < .05 level) than the difference between white graduates and white dropouts, but approaches significance at p < .09. At age 22: African American graduates—dropouts significantly different at p < .001; white graduates—dropouts show no statistical difference; the difference between African American graduates and African American dropouts is statistically different than the difference between white graduates and white dropouts at p < .01

Results in Fig. 2 show a striking pattern. At each age, African American high school graduates exhibit a steady percentage point increase in the proportion of weeks worked full time compared to African American dropouts. Among white high school graduates and dropouts, by contrast, the percentage point difference in weeks worked full time peaks at age 20 (11 percentage points) and declines at ages 21 and 22. At age 22 African American high school graduates and dropouts have a 16 percentage point difference in the percent of weeks worked. If, for example, an African American high school graduate has worked full time for the entire year, an African American peer without a high school diploma would have gone two full months that year without full time employment.

The difference in proportion of weeks worked among African Americans is statistically significant (p < .05) at each age. Among white high school graduates and dropouts the differences are statistically significant at each age except 22. We examined whether the difference between high school graduates and dropouts varied by race, by testing whether the percentage point difference between African American graduates and dropouts was statistically different than the percentage point difference between white high school graduates and dropouts. The difference between African American and white youth does not reach statistical significance until age 22 when African Americans have a 16 percentage point difference and white dropouts and graduates have converged to only a 5 percentage point difference (p < .01). Despite the statistical insignificance of other ages, Fig. 2 suggests distinct racial patterns leading up to the significant difference at age 22. For African American youth, a high school diploma becomes increasingly important over time. Among white youth, a high school diploma appears to decrease in significance (relative to white dropouts)—at least with respect to the proportion of weeks worked full time.

Conclusion and implications

Among youth who do not pursue post-secondary education, having a high school diploma means significantly more employment during the initial years of the transition to adulthood than it does without one. This finding, from a cohort of youth who turned 18 between 1998 and 2001, supports other research showing higher overall earnings for high school graduates compared to high school dropouts (Goldin and Katz 2008). What our analyses also show, however, is that African American high school graduates work about as many weeks as white high school dropouts, and significantly fewer weeks than white high school graduates at age 18, holding family and youth characteristics constant. They also take significantly longer than white high school graduates, and about as long as white high school dropouts to attain 6 consecutive months of full time work after age 18, with family and youth characteristics held constant. At age 22, African American high school graduates work full time roughly as many weeks as both white high school graduates and white dropouts—holding other family and youth characteristics constant.

One interpretation of these findings is that a high school diploma means very little for African Americans compared to whites during the transition to adulthood since they work about as much as white youth who do not complete high school. The interpretation is valid until one considers the findings for African American dropouts. African American youth who do not complete high school work significantly less than African American graduates, white graduates, and white dropouts at each age. They take 240 % longer than white graduates to secure 6 consecutive months of full time employment after age 18, work 17 percentage points fewer weeks at age 18, and 19 percentage points fewer weeks at age 22.

The results suggest that while a high school diploma leads to significantly more weeks of employment for white high school graduates compared to African American high school graduates, having the diploma may be more important for African Americans. Other authors have also noted that education can bring (relative) larger returns to African Americans (Lang 2007; Lang and Lehman 2011). When we examine within-race differences between graduates and dropouts we find the differences in employment between African American high school graduates and dropouts are significantly larger than the differences between white high school graduates and dropouts. In fact, the difference between African American high school graduates and dropouts increases steadily from an 8 to a 16 percentage point difference between ages 18 and 22, while the differences rise and then fall by age 22 for white youth—never topping 11 %.

It appears that over time having a high school diploma becomes more important to African American employment as the gap between high school graduates and dropouts widens. Some explanations could be that African American dropouts are not acquiring early work experience that employers use to base hiring decisions on. While we control for work experience before age 18, it is clear that African American dropouts acquire less work experience than their peers after age 18. Other likely explanations, however, include fewer job networks, discrimination, and the concomitant impact that discrimination could have on the role of signals in the labor market. If for instance white employers are more reliant on signals to assess the quality of African American applicants who they may be less likely to hire, a signal from a trusted institution (like the school system) may have more enduring relevance.

An important unobserved factor that may explain the increasing advantage that graduates have in finding employment is the extent of their networks and social connections. African American dropouts may have fewer networks that can connect them to jobs, so that initial lapses into unemployment are harder to recover from for dropouts.

This study considers the determinants of gaps in employment between high school graduates and dropouts, and between African American and white youth. However, substantial gaps not only exist in employment, but in the decision to enter the labor force in the first place (Holzer 2009). Theories associated with human capital, discrimination, and signaling presuppose that youth are looking for work, which may not be the case. Holzer (2009) suggests that declining job opportunities at acceptable wages and increasingly remunerative illegal activities have played an important role in the withdrawal of young African American men from the labor market. Wilson (1996) similarly points out that this withdrawal is not exogenous to labor market conditions. The lack of employment opportunities for African Americans with low education levels, and the problems associated with discrimination and signaling play an important role in the decision to withdraw from the labor market.

Deriving policy recommendations from these findings is somewhat challenging, as it always is in discussions about the returns to education. The decision to get an education is a highly endogenous decision. Less academically successful or skilled students are less likely to get a diploma; and simply increasing the rate at which diplomas are awarded to these students may not provide them with the aptitude or skills necessary to increase employment productivity. Nevertheless, the persistence of employment advantages for graduates suggests that some combination of benefits is being provided at schools meriting robust dropout prevention policies. In a world of lower dropout rates the endogeneity concerns that plague the literature on the returns to education may justify hesitancy towards a one-size-fits-all approach to education policy. However, one of the advantages of this study is that it differentiates between the premium offered by a high school diploma for whites and the same premium for African Americans. It is evident that the employment advantages enjoyed by African American graduates over African American dropouts indicate the benefits of high school completion, even among youth who historically fare less favorably in the labor market.

We find that the traditional attention on disparities between African American youth and white youth obscures the differences among African Americans in employment rates. The fact that African American graduates do no better in the job market than white dropouts is disconcerting; the fact that the advantages of a high school diploma are much greater for African American youth than for white youth suggest a policy lever—dropout prevention policies—may be exploited. However, while dropout prevention may address a major source of employment inequality, African American youth who have already dropped out and disadvantaged high school graduates would not benefit.

Some disparities experienced by African American youth are left unexplained by our analysis. A large literature exists that attempts to identify the source of this bedrock finding of persistent disparity between African Americans and whites. Our contribution to this literature is to point out that the difference in the value of a high school diploma between African Americans and whites may corroborate the theory that employers rely on signals and statistical discrimination in hiring workers, but that they rely on those signals asymmetrically. Signals like a high school diploma are more valuable for African American youth than for whites, perhaps because employers are more reliant on signaling in deciding whether to hire an African American applicant. While our findings may bolster the case for signaling theory, further research will continue to be essential to untangle the complex causes of racial disparities in employment in the United States. This paper suggests that signaling in the labor market may work asymmetrically across racial groups; further research could also explore whether asymmetric signaling helps to explain differences in employment outcomes between African American males and females.

Footnotes
1

One element of Heckman’s (1998) criticism is that employers and audit-study designers may be attuned to different observable characteristics—and while audit studies match fake job applicants on one set (e.g., résumé, attire, comportment), employers may observe and focus on additional features, still unrelated to race, that study-designers may not know or match on. Therefore, the job applicants are not truly identical from an employer’s perspective.

 
2

Youth who graduate from high school after their 18th birthday may be at a disadvantage relative to other youth in the employment outcome variables, which are measured from the 18th birthday. To determine whether this effect is significant, we performed a sensitivity analysis by adding a control variable for the number of months after age 18 that a youth graduated from high school. This variable was set to zero for drop outs and youth who graduated before they turned 18 (since these youth would not be in school when their employment outcomes were measured). Qualitatively, the results for all of the sensitivity analyses were identical to the analyses presented here, suggesting that the birth dates of the youth had no impact on the analyses. In some cases the estimated coefficients changed by adding this control variable, but none of the conclusions of the paper, and none of the disparities between white graduates, white drop outs, African American graduates, and African American drop outs reported in the paper changed in the sensitivity analyses.

 
3

As a robustness check, the regression models were also conducted using a censored normal model and a negative binomial model. African American graduates and dropouts, as well as white dropouts performed substantially worse than white graduates in all of these models, relative to the OLS models presented in the paper. In that sense, the results presented in this paper represent the most conservative results produced out of all the model specifications that were conducted. Censored normal and negative binomial results are available from the authors on request.

 
4

As a robustness check, the accelerated failure time models were also conducted using the log-logistic distribution. None of the qualitative results changed as a result, although estimated time to failure relative to the reference group of white high school graduates was somewhat longer. Therefore, the Weibull distribution results reported here can be considered conservative estimates of racial disparities in time until employment. Log-logistic distribution results are available from the authors on request.

 

Copyright information

© Springer Science+Business Media New York 2012