Open Access

Is it worth it? Postsecondary education and labor market outcomes for the disadvantaged

IZA Journal of Labor Policy20154:1

DOI: 10.1186/s40173-014-0027-0

Received: 27 August 2014

Accepted: 29 November 2014

Published: 23 January 2015

Abstract

In this paper we examine a range of postsecondary education and labor market outcomes, with a particular focus on minorities and/or disadvantaged workers. We use administrative data from the state of Florida, where postsecondary student records have been linked to UI earnings data and also to secondary education records. Our main findings can be summarized as follows: 1) Gaps in secondary school achievement can account for a large portion of the variation in postsecondary attainment and labor market outcomes between the disadvantaged and other students, but meaningful gaps also exist within achievement groups, and 2) Earnings of the disadvantaged are hurt by low completion rates in postsecondary programs, poor performance during college, and not choosing high-earning fields. In particular, significant labor market premia can be earned in a variety of more technical certificate and Associate (AA) programs, even for those with weak earlier academic performance, but instead many disadvantaged (and other) students choose general humanities programs at the AA (and even the BA level) with low completion rates and low compensation afterwards. A range of policies and practices might be used to improve student choices as well as their completion rates and earnings.

Keywords

Disadvantaged Postsecondary Earnings Achievement Completion Associate Degree Bachelor's Degree

1 Introduction

It is, by now, well-known that rewards to college degrees (especially at the BA level) have grown quite large in the U.S. labor market. Indeed, for young people growing up in disadvantaged families, obtaining a college degree is the surest way to achieve upward mobility for themselves and their families (Isaacs 2007; Haskins et al. 2009). Thus, the incentives for low-income young people to pursue higher education have become very strong. Even if only an associate degree or a certificate is achieved, the labor market rewards for young people remain quite substantial (Holzer and Dunlop 2013).

Yet, in spite of these incentives, young economically disadvantaged students lag substantially behind their middle- and upper-income counterparts in achieving postsecondary credentials, and minorities continue to lag behind white students. While rates of college enrollment have risen for all groups in recent years, college completion rates, especially among minorities and the disadvantaged, remain low (Bound et al. 2009; Holzer and Dunlop 2013), and gaps in postsecondary attainment between income groups in the U.S. have grown in recent decades (Bailey and Dynarski 2011). In addition, while the average value of a college degree in the labor market is high, the concentration of young people in high-earning fields (like science, technology, engineering, and math, or STEM) is not as high as we might expect, especially given the meaningful differences in earnings we observe across fields. In particular, women, minorities and the poor are less likely to earn a credential in these high-earning fields.

Why are disadvantaged and/or minority students less likely than their white/middle-class counterparts to earn postsecondary credentials? In part, this is because of the academic achievement gap between these students and their more advantaged peers. This gap emerges early in life and tends to become larger as students progress through school (Reardon 2011). But even adjusting for prior achievement, disadvantaged young people lag behind others in college completion and attainment.

If the achievement gap does not fully explain this difference, what does? The research literature has identified information gaps, college costs and lack of full-time attendance as some of the factors that impair the success rates of disadvantaged young people in college (Bound et al. 2009; Hoxby and Turner 2013). Because of these factors, young disadvantaged students attend lower-quality colleges and universities, with lower completion rates, and their completion rates within the same institutions are lower than those of more advantaged students as well.

But many questions remain unanswered about the relative magnitudes and the interaction of factors that contribute to the lack of postsecondary success for disadvantaged students. For example, how important is disadvantaged students’ choice of major at each level of education? Adjusting for high school achievement, by how much does their performance in college – as measured by courses taken, grades attained, credits earned, and ultimately program completion - lag behind the postsecondary performances of students who are not disadvantaged? And how much do these factors account for their lower labor market earnings, as opposed to other barriers that impede the accumulation of valuable labor market experience (Johnson and Neal, 1998)?

Answers to these questions are important if we want to design effective programs and policies to better assist disadvantaged students in their college experiences; and such answers require detailed longitudinal micro data on students, their educational institutions and experiences, and labor market outcomes. While some such information is available in existing longitudinal survey datasets on young people – such as the National Educational Longitudinal Study (NELS), High School and Beyond (HSB) or the National Longitudinal Survey of Youth (NSLY) - administrative data on students provides more detailed information on every course taken and on all academic outcomes achieved for every student who ever attended a public school in the relevant years. Until recently such data have not existed at the state level, but in several states these data are now becoming available. This enables researchers to address previously unexplored questions about the experience and outcomes associated with disadvantaged students in a number of contexts1.

In this paper, we use administrative data from the state of Florida to analyze the college and labor market experiences of two cohorts of young people. We are able to extend the current literature on the returns to college credentials in several important ways. First, the students in our data graduated high school between 2000 and 2002, making them a more recent cohort than many previous studies. Second, our large sample size, over 210,000 students, allows us to measure heterogeneous effects with precision. Third, unlike many other recent studies, we also have access to secondary school data, so we can control for earlier achievement. Finally, this paper focuses on disadvantaged students in particular, a large and growing fraction of the U.S. postsecondary market.

Below, we describe our data and analysis in Part II of the paper, present the empirical results in Part III, and present our conclusions and their general implications for further research and policy in Part IV.

2 Data and analysis plan

2.1 Data overview

To create our sample, we merge together three large administrative student-level data sets from the state of Florida: secondary school data, postsecondary data, and unemployment earnings data, the latter of which provides earnings information for nearly all Florida residents2. Linked together by a unique individual identifier, we are able to follow students from eighth grade through college (and graduate school) and their entry into the labor force, provided the student does not leave the state of Florida. Our data provide large sample sizes unavailable in national surveys, offer rich information to account for selection based on ability (e.g. Lovenheim and Reynolds 2011), and provide detailed information on several important outcome measures.

The data for this paper follow two cohorts of students. The first wave comprises all students who began 10th grade at a Florida public school in 1997–98. The second wave consists of all students who began 8th grade in 1997–1998. The data set contains observations as recent as 2011–12, so we observe 10–12 years of postsecondary and labor market outcomes.

The secondary student-level data include student demographic information (race, gender, and limited English proficiency status), and we use eligibility for free- or reduced-price lunch (FRL) as a measure of family income. In addition, we have data on courses taken, course grades, GPA, and standardized test scores (such as the Florida Comprehensive Assessment Test, or FCAT). These serve as important controls and allow for analysis of heterogeneous treatment effects.

Our rich postsecondary information comes from public colleges and universities in Florida and includes date and institution of each enrollment instance, courses taken at each institution and grades received, cumulative credits earned, field of study, and degree attainment. We utilize this data to measure both intermediate postsecondary outcomes (such as major choice) and terminal postsecondary outcomes (such as highest degree earned).

Finally, we merge our data with quarterly wage information from employer reports to Florida’s Unemployment Insurance (UI) office. These data are collected during high school and for five years after students leave their last educational institution.

The main drawback of our data is their limitation to a single state, which has implications for both generalizability to other states and attrition (since individuals leaving the state are not observed). For generalizability, this drawback is somewhat mitigated by Florida's large and diverse makeup. According to the 2012 Digest of Education Statistics, Florida ranked as the fourth largest state in terms of the number of graduates from public high schools. In addition, 22 percent of its high school graduates were black and 23 percent Hispanic, compared to national averages of 15 and 16 percent, respectively. Given this paper’s focus on disadvantaged students, using data from such a large and highly diverse state increases sample sizes and the precision of our estimates for these groups.

Although the data are limited to students who attend in-state public institutions, Florida has a relatively low out-migration rate for college-going. Specifically, only 10 percent of Floridian first-time degree or certificate-seeking undergraduates attend college out of state3. In addition, about 85% percent of first-time degree-seeking students attend a public, rather than private, institution in Florida4. Therefore, we likely do not lose a large subset of our sample to private or out-of-state colleges. However, students who enroll in out-of-state or private colleges are more likely to come from higher-income families, are higher-performing students, and may be attending higher-quality institutions. The potential downward bias in some of our estimates of enrollment, completion, and earnings due to missing student observations needs to be kept in mind as we review our results.

In addition, as many as half of all Florida higher education enrollees do not subsequently appear in the labor market data. This could be either because they have moved out of state for work, cannot find employment, or have chosen not to enter the labor market. The first of these explanations might generate some under-representation of high-earning groups in our data, while the others might instead generate some under-representation of low earners. Previous studies have found that 30-50% of males age 25–45 work in a state other than their birth state, depending on their level of education, which is broadly consistent with the out-migration we observe5. In addition, our own calculations using the American Community Survey find similar rates of mobility of college graduates who were born in Florida. Overall, rates of labor market participation appear lowest for those with the weakest observed skills in their high school years6.

A final limitation of the sample is its emphasis on traditional students: those whose postsecondary studies begin relatively soon after high school graduation. For example, if a student were to go back to college outside the 10–12 year window for which we have data, this enrollment would not be captured. Thus, we miss out on this important component of the community college-going population7.

We construct several variables from our administrative data. For example, for students who complete a degree, identifying degree earned and field of study is straightforward. However, students who enroll in postsecondary education but do not complete a degree could have had many fields of study throughout the course of their enrollment. Reasonable choices for “field of study” include first declared major, last declared major, or most frequently-observed major (by number of terms). In this paper, for those students who do not earn a degree, we consider their field of study to be their final declared major before exiting postsecondary schooling.

Finally, we count each quarter that an individual appears in the UI data as one quarter of work experience. When calculating tenure, we count each quarter of employment with a given employer, whether or not that employment is continuous8. When examining labor market outcomes, we only consider observations following each individual’s final term of postsecondary enrollment and only include students 18 years of age and older. Only those quarters with positive earnings are included in our computations9.

2.2 Analysis plan

A number of recent studies have examined labor market returns to various fields and credentials (Jepsen et al. 2014 (Kentucky); Bahr et al. 2014 (Michigan); Bahr et al. 2014 (California); Belfield et al. 2014 (North Carolina)). However, the data underlying these studies lack information about students’ high school experiences. In order to estimate postsecondary returns, these studies compare pre-schooling earnings with post-schooling earnings in a fixed effect framework and attribute the post-schooling earnings gain to the effect of postsecondary training. However, a recent study by Xu and Jaggers (2014) provides evidence that pre-schooling wages of younger students understate their actual earning potential and that the assumptions underlying fixed effects estimation may not hold. Therefore, we control for a student’s academic performance in high school as a measure of their earning potential.

We begin our analysis by presenting summary data on differences in higher education and labor market outcomes between race/gender groups, and those who do and do not qualify for FRL. After establishing the basic facts on outcomes that need to be explained, we divide the students into quartiles based on their high school FCAT scores and explore differences between these quartiles – which likely reflect their differences in early achievement – versus those differences within quartiles, which cannot be attributed to achievement gaps10.

We then present results from regression equations of the following general form:
$$ {\mathrm{lnEARN}}_{\mathrm{i}\mathrm{kt}} = \mathrm{f}\left({\mathrm{ED}}_{\mathrm{i}},\ {\mathrm{X}}_{\mathrm{i}},\ {\mathrm{ACHIEVE}}_{\mathrm{i}},\ {\mathrm{EXP}}_{\mathrm{i}},\ {{\mathrm{EXP}}_{\mathrm{i}}}^2,\ {\mathrm{TEN}}_{\mathrm{i}},\ {{\mathrm{TEN}}_{\mathrm{i}}}^2,\ {\mathrm{COHORT}}_{\mathrm{k}},\ {\mathrm{TIME}}_{\mathrm{t}}\right) + {\mathrm{u}}_{\mathrm{i}\mathrm{kt},} $$
(1)
where lnEARN denotes the natural log of quarterly earnings; ED denotes the highest level of education completed (either high school, a certificate, an AA degree, a BA degree, or higher); X measures race/ethnicity, gender, and family background;11 ACHIEVE measures high school achievement (FCAT math score);12 EXP and TEN reflect quarters of total labor market experience and job tenure with their current firm respectively (entering the equation in both linear and quadratic form); and COHORT and TIME denote cohort and year/quarter dummies. The individual person, cohort, and year/quarter are denoted respectively by the subscripts i, k, and t. Missing values in achievement measures are measured as zeroes along with a “1” for a missing value dummy variable.

We use our measure of high school achievement to correct for selection into higher education, which, as discussed above, is unobservable in most studies. For the reasons noted above, we regard fixed effects models as not appropriate in this setting and thus do not estimate them.

In some versions of the estimated equations, we control for the number of postsecondary credits earned if the student did not complete the degree program in which they had enrolled. In other equations, we add measures of postsecondary achievement – such as college GPA and the number of credits earned in math or science courses – to see the extent to which such achievement is rewarded in the job market. Returns to different fields of study at different levels of higher education are estimated in separate equations as well. Finally, we separate out those achieving different types of associate degrees (such as Associate in Arts v. Associate in Applied Science, or AA v. AAS) to see the extent to which the kind of degree achieved affects subsequent labor market earnings as well.

3 Empirical results

3.1 Summary results

Table 1 presents our estimates of higher education and labor market outcomes for Florida students in our sample. These outcomes include quarterly earnings as well as observed high school completion rates; higher educational attainment and enrollment rates, conditional on completing high school; and higher education completion rates, conditional on enrollment. These results appear for all students and workers in part A, and then separately for race/gender groups and for FRL/gender groups respectively in parts B and C.
Table 1

Education and labor market outcomes: summary statistics

A. All students and workers

   

Quarterly earnings

$5,227.36

     
 

(4572.22)

     

 N

3,975,013

     

Educational attainment

      

 HS

0.693

     

 Voc/Cert

0.044

     

 AA

0.159

     

 BA or above

0.193

     

 N

393,213

     

Enrollment

      

 Voc/Cert

0.106

     

 AA

0.529

     

 BA

0.283

     

Completion

      

 Voc/Cert

0.419

     

 AA

0.300

     

 BA

0.592

     

B. By race/gender

 

White

Black

Hispanic

 

Male

Female

Male

Female

Male

Female

Quarterly earnings

$5,746.43

5,245.16

4,237.60

4,140.88

5,654.21

5,473.62

 

(4967.67)

(4575.25)

(3779.69)

(3625.38)

(4638.84)

(4332.01)

 N

1,243,071

906,996

528,548

463,994

415,221

320,809

Educational attainment

      

 HS

0.716

0.774

0.531

0.664

0.619

0.700

 Voc/Cert

0.063

0.037

0.034

0.040

0.048

0.025

 AA

0.137

0.209

0.066

0.132

0.135

0.206

 BA or above

0.168

0.266

0.075

0.170

0.125

0.217

Enrollment

      

 Voc/Cert

0.113

0.082

0.121

0.147

0.116

0.085

 AA

0.459

0.561

0.433

0.594

0.556

0.640

 BA

0.255

0.347

0.157

0.265

0.224

0.322

Completion

      

 Voc/Cert

0.555

0.445

0.281

0.274

0.414

0.289

 AA

0.299

0.373

0.153

0.222

0.243

0.322

 BA

0.580

0.658

0.422

0.549

0.494

0.573

C. By FRL/gender

 

Non-FRL

FRL

  
 

Male

Female

Male

Female

  

Quarterly earnings

$5,943.05

5,610.97

4,740.15

4,236.17

  
 

(5138.77)

(4755.39)

(3965.59)

(3555.97)

  

 N

1,241,573

968,371

865,102

698,666

  

Educational attainment

      

 HS

0.744

0.812

0.589

0.671

  

 Voc/Cert

0.059

0.035

0.044

0.037

  

 AA

0.145

0.217

0.087

0.147

  

 BA or above

0.187

0.295

0.065

0.126

  

Enrollment

      

 Voc/Cert

0.113

0.085

0.118

0.120

  

 AA

0.501

0.590

0.448

0.588

  

 BA

0.288

0.389

0.139

0.211

  

Completion

      

 Voc/Cert

0.521

0.408

0.377

0.310

  

 AA

0.290

0.368

0.194

0.249

  

 BA

0.572

0.648

0.421

0.519

  

Notes: The number of quarterly earning observations are the number of labor market quarters observed while educational outcomes are calculated out of unique students. Standard deviations for continuous variables are in parentheses below their corresponding means. All postsecondary attainment and enrollment are conditional upon HS graduation. Completion for a degree level is conditional upon enrollment in that degree.

The results in part A show that quarterly earnings for all workers in our sample average about $5,200 and have considerable variation, though this estimate is likely biased by out-of-state migration of higher-achieving students and workers on the high end of the wage distribution, and by labor force nonparticipation on the low end13. Just over two-thirds of all students complete high school – a rate roughly consistent with computations using administrative data from other states in this period14.

In our sample, 11, 53, and 28% of high school graduates enroll in vocational certificate, AA, and BA programs, respectively. As shown in part B, both Hispanic males and females are relatively more likely to enroll in AA programs, with Hispanic females being the most likely to pursue an AA degree (64%) and black males being the least likely (43%). On the other hand, white students are relatively more likely to enroll in a BA program.

The data also indicate that 4, 16, and 19% of high school graduates obtain vocational certificates, AA, and BA degrees, respectively, in public institutions in Florida. These numbers are substantially lower than the share of students who enroll, due to low completion rates, especially for those enrolled in AA programs. Large differences in completion rates are apparent across racial/gender groups. For every 100 white male students who enroll in an AA program, about 30 will earn an associate’s degree, compared to 15 for black males and 24 for Hispanic males. As shown in Table 1C, completion rates are similarly lower for FRL students relative to non-FRL students.

It should be noted that the BA enrollment and completion rates are also likely downward biased by the out-migration of higher-achieving and higher-income students in these data, as noted above, but the observed AA rates are likely much less downward-biased by these factors. The fact that completion rates are somewhat higher in vocational certificate programs than in AA programs, though overall enrollments are much lower, is notable as well.

Overall in Table 1, we find numerous results, many of which have been observed elsewhere (e.g., Holzer and Dunlop, op. cit):
  • Whites earn more than blacks and have higher rates of educational attainment, enrollment and completion than blacks or Hispanics;

  • The poor, (as measured by FRL) also have lower rates of attainment, enrollment, and completion; and

  • Educational outcomes are generally higher for females in each racial or income group though their earnings are lower.

We also note that educational attainment and labor market outcomes for Hispanics are consistently better than those for blacks; this partly, but not entirely, reflects the presence of higher-achieving and higher-earning Cubans among the latter (Borjas 1987). However, outcomes tend to be better for non-Cuban Hispanics too15. Higher education and labor market outcomes of those from FRL households are generally similar to, or a bit better than, those observed for blacks in Florida.

Two other outcomes are notable as well: first, males are generally more likely than females to enroll in and complete vocational programs; and second, black males lag behind black females in enrollment and completion by even more than those of other racial groups, and they earn only marginally more than the females in the job market. For instance, the rate of both BA and AA attainment for black females is more than twice as high as that of black males. While 17% of black females earn a bachelor’s degree, less than 8% of black males do. It is also important to note that the general underrepresentation of black (and/or) low-income men that we commonly find in survey data might create less bias in these administrative data, but perhaps some bias still exists16.

Table 2 presents some additional estimates of intermediate higher education and labor market outcomes. These include our measures of high school achievement – GPA and math FCAT – as well as somewhat similar measures of postsecondary achievement – postsecondary GPA, credits earned (among all students, whether or not they have completed their programs), and numbers of math or science credits. Intermediate outcomes for the labor market – work experience and job tenure – appear as well. Again, estimates of outcomes appear for the entire sample of students and workers in part A, and for race/gender and FRL/gender groups in parts B and C of the table, respectively.
Table 2

Intermediate education and labor market outcomes: summary statistics

A. All students/workers

Intermediate HS outcomes

      

 10th grade FCAT math score

0.000

     
 

(1.000)

     

 HS GPA

2.208

     
 

(0.982)

     

Intermediate postsecondary outcomes

      

 Postsecondary GPA

2.271

     
 

(1.083)

     

 Total postsecondary credits

71.450

     
 

(61.321)

     

 Postsecondary math credits

7.730

     
 

(8.207)

     

 Postsecondary science credits

11.016

     
 

(17.469)

     

Labor market inputs

      

 Work experience

18.886

     
 

(11.272)

     

 Tenure

6.206

     
 

(6.897)

     

B. By race/gender

 

White

Black

Hispanic

 

Male

Female

Male

Female

Male

Female

Intermediate HS outcomes

      

 10th grade FCAT math score

0.287

0.264

−0.691

−0.561

−0.149

−0.175

 

(0.922)

(0.822)

(1.074)

(0.966)

(1.006)

(0.913)

 HS GPA

2.188

2.533

1.804

2.087

1.945

2.263

 

(0.994)

(0.972)

(0.891)

(0.912)

(0.920)

(0.912)

Intermediate postsecondary outcomes

      

 Postsecondary GPA

2.266

2.568

1.718

1.967

2.062

2.301

 

(1.102)

(0.995)

(1.103)

(1.048)

(1.089)

(1.013)

 Total postsecondary credits

68.515

79.069

50.597

68.157

63.976

77.191

 

(60.08)

(60.723)

(57.608)

(63.531)

(58.351)

(60.534)

 Postsecondary math credits

7.657

7.732

6.055

7.347

8.261

8.800

 

(8.682)

(7.189)

(8.573)

(8.062)

(9.153)

(8.104)

 Postsecondary science credits

9.331

12.999

6.53

11.4

8.518

12.219

 

(15.527)

(18.504)

(13.53)

(18.211)

(14.853)

(18.246)

Labor market inputs

      

 Work experience

19.451

19.12

17.703

18.401

19.027

18.955

 

(11.397)

(10.881)

(11.567)

(11.214)

(11.536)

(11.109)

 Tenure

6.500

6.357

5.616

5.723

6.174

6.421

 

(7.274)

(6.794)

(6.696)

(6.437)

(6.882)

(6.747)

C. By FRL/gender

 

Non-FRL

FRL

  
 

Male

Female

Male

Female

  

Intermediate HS outcomes

      

 10th grade FCAT math score

0.242

0.220

−0.381

−0.398

  
 

(0.954)

(0.856)

(1.035)

(0.928)

  

 HS GPA

2.286

2.628

1.805

2.088

  
 

(0.936)

(0.901)

(0.909)

(0.931)

  

Intermediate postsecondary outcomes

      

 Postsecondary GPA

2.240

2.531

1.877

2.072

  
 

(1.075)

(0.970)

(1.153)

(1.090)

  

 Total postsecondary credits

71.219

83.558

50.689

62.575

  
 

(61.049)

(61.769)

(54.998)

(59.091)

  

 Postsecondary math credits

8.097

8.253

6.496

7.266

  
 

(8.926)

(7.509)

(8.692)

(7.932)

  

 Postsecondary science credits

9.927

14.052

6.620

10.037

  
 

(16.161)

(19.571)

(13.549)

(16.577)

  

Labor market inputs

      

 Work experience

19.388

19.381

18.432

18.34

  
 

(11.475)

(10.941)

(11.441)

(11.089)

  

 Tenure

6.536

6.438

5.757

5.843

  
 

(7.256)

(6.782)

(6.661)

(6.509)

  

Note: Intermediate HS and postsecondary outcomes are calculated out of numbers of unique students, while labor market inputs are calculated out of labor market quarters.

The results in part A show high school and postsecondary GPAs that average about 2.2. About a fourth of all postsecondary credits earned are in math or science, and most workers have accumulated over 4 years of work experience on average over our sample period, with a third of these quarters being with their current/most recent employer.

Comparing these outcomes across race/gender and FRL/gender groups generates findings similar to those in Table 1. Namely, males generally have more labor market experience or tenure than females in each group (with black students being the exception) but lower GPAs and postsecondary credits earned. Results for FCAT math scores are more mixed – with higher scores among males than females in most groups but lower scores among black males than females. Once again, minorities score lower than whites on most academic measures and FRL students score lower than non-FRL. The average white male accumulates about 68 postsecondary credits, compared to 51 for black males and 64 for Hispanic males. In addition, Hispanics outperform blacks (and FRL) on academic measures, and even accumulate nearly as much work experience as whites. Black males lag behind their female counterparts in educational outcomes by more than we observe for other groups. For example, the average black female earns 68 postsecondary credits with a postsecondary GPA of about 2.0, while the average black male earns 51 credits with an average GPA of 1.7.

3.2 Selection by high school achievement

Undoubtedly, young people with different levels of academic achievement self-select into different higher education programs, institutions, and fields of study and differ in terms of their labor market participation and occupational choices. We investigate the extent to which differences in academic or labor market performance reflect differential selection based on students’ ability and early academic achievement.
Table 3

Education and labor market outcomes: by hs achievement quartile

A. All students/workers

 

Quartile 1

Quartile 2

Quartile 3

Quartile 4

        

Quarterly earnings

$4,383.75

5,227.19

6,159.69

8,072.06

        
 

(3411.79)

(3955.62)

(4765.38)

(6445.16)

        

Educational attainment

            

 Voc/Cert

0.045

0.054

0.054

0.035

        

 AA

0.082

0.154

0.220

0.224

        

 BA or above

0.041

0.104

0.226

0.453

        

Enrollment

            

 Voc/Cert

0.143

0.119

0.100

0.060

        

 AA

0.536

0.590

0.600

0.484

        

 BA

0.089

0.190

0.351

0.580

        

Completion

            

 Voc/Cert

0.317

0.456

0.542

0.585

        

 AA

0.153

0.261

0.367

0.462

        

 BA

0.425

0.500

0.577

0.656

        

Intermediate HS outcomes

            

 10th grade FCAT Math Score

−1.286

−0.194

0.368

1.131

        
 

(0.823)

(0.170)

(0.166)

(0.404)

        

 HS GPA

1.977

2.342

2.680

3.159

        
 

(0.679)

(0.647)

(0.640)

(0.588)

        

Intermediate postsecondary outcomes

        

 Postsecondary GPA

1.734

2.120

2.427

2.841

        
 

(1.085)

(0.995)

(0.930)

(0.810)

        

 Total postsecondary credits

43.542

62.603

80.698

101.912

        
 

(49.200)

(56.430)

(59.703)

(61.394)

        

 Postsecondary math credits

5.401

7.957

8.893

9.481

        
 

(7.706)

(8.079)

(7.736)

(8.558)

        

 Postsecondary science credits

5.476

8.908

12.638

17.412

        
 

(10.805)

(14.077)

(17.949)

(22.212)

        

Labor market inputs

        

 Work experience

18.997

19.323

19.486

18.843

        
 

(11.453)

(11.063)

(10.758)

(10.032)

        

 Tenure

6.187

6.619

6.778

6.831

        
 

(6.944)

(7.126)

(7.002)

(6.669)

        

B. By race

 

Quartile 1

Quartile 2

Quartile 3

Quartile 4

 

White

Black

Hispanic

White

Black

Hispanic

White

Black

Hispanic

White

Black

Hispanic

Quarterly earnings

$4,455.57

4,036.84

4,987.81

5,190.27

4,823.83

5,791.99

6,085.60

5,687.52

6,711.49

7,934.70

7,793.08

8,626.83

 

(3528.71)

(3187.7)

(3553.55)

(3967.44)

(3684.96)

(4138.44)

(4703.92)

(4516.54)

(4982.39)

(6312.65)

(6599.3)

(6615.71)

Educational attainment

            

 Voc/Cert

0.058

0.041

0.032

0.063

0.047

0.042

0.061

0.038

0.044

0.038

0.020

0.029

 AA

0.069

0.075

0.118

0.149

0.133

0.186

0.226

0.162

0.237

0.232

0.166

0.211

 BA

0.030

0.045

0.049

0.084

0.133

0.109

0.207

0.280

0.240

0.446

0.470

0.444

Enrollment

            

 Voc/Cert

0.135

0.159

0.120

0.127

0.118

0.102

0.105

0.093

0.089

0.063

0.047

0.054

 AA

0.442

0.559

0.629

0.556

0.582

0.674

0.598

0.542

0.644

0.494

0.413

0.481

 BA

0.060

0.100

0.111

0.148

0.249

0.210

0.314

0.444

0.386

0.566

0.637

0.600

Completion

            

 Voc/Cert

0.432

0.260

0.268

0.499

0.398

0.411

0.578

0.410

0.499

0.603

0.423

0.535

 AA

0.155

0.134

0.188

0.267

0.228

0.276

0.378

0.298

0.368

0.469

0.402

0.439

 BA

0.474

0.414

0.394

0.539

0.478

0.464

0.595

0.550

0.549

0.665

0.590

0.622

Intermediate HS outcomes

            

 10th grade FCAT Math Score

−1.175

−1.385

−1.260

−0.181

−0.217

−0.199

0.377

0.333

0.358

1.141

1.015

1.081

 

(0.762)

(0.869)

(0.796)

(0.169)

(0.17)

(0.171)

(0.165)

(0.163)

(0.165)

(0.409)

(0.305)

(0.349)

 HS GPA

1.998

1.962

1.968

2.340

2.364

2.308

2.688

2.696

2.616

3.176

3.046

3.052

 

(0.722)

(0.653)

(0.668)

(0.665)

(0.617)

(0.637)

(0.654)

(0.601)

(0.618)

(0.594)

(0.550)

(0.573)

Intermediate postsecondary outcomes

            

 Postsecondary GPA

1.849

1.646

1.767

2.179

1.999

2.120

2.463

2.261

2.423

2.857

2.653

2.788

 

(1.142)

(1.047)

(1.06)

(1.029)

(0.936)

(0.981)

(0.941)

(0.890)

(0.911)

(0.816)

(0.786)

(0.792)

 Total postsecondary credits

38.880

44.080

49.196

56.596

69.716

65.880

76.475

91.804

84.474

100.025

114.108

102.806

 

(44.219)

(50.849)

(51.569)

(52.473)

(61.426)

(56.218)

(57.904)

(65.152)

(59.144)

(60.809)

(66.257)

(59.791)

 Postsecondary math credits

4.701

5.302

6.586

7.144

8.458

9.052

8.312

9.727

10.021

9.149

10.817

10.337

 

(6.97)

(7.739)

(8.468)

(7.556)

(8.321)

(8.657)

(7.341)

(8.491)

(8.068)

(8.337)

(9.398)

(8.614)

 Postsecondary science credits

4.585

5.752

5.992

7.692

10.674

8.908

11.485

15.867

12.958

16.469

21.914

18.003

 

(9.041)

(11.297)

(11.547)

(12.14)

(16.431)

(13.651)

(16.444)

(20.978)

(18.201)

(21.238)

(26.452)

(22.569)

Labor market inputs

            

 Work experience

19.681

18.525

18.973

19.707

18.560

19.439

19.804

18.407

19.277

19.163

17.461

18.082

 

(11.565)

(11.373)

(11.43)

(11.138)

(10.843)

(11.129)

(10.826)

(10.495)

(10.672)

(10.084)

(9.767)

(9.833)

 Tenure

6.417

5.968

6.286

6.867

6.061

6.692

6.970

5.996

6.672

6.952

6.044

6.550

 

(7.292)

(6.736)

(6.793)

(7.445)

(6.529)

(6.989)

(7.212)

(6.312)

(6.679)

(6.797)

(5.927)

(6.257)

C. By FRL status

 

Quartile 1

Quartile 2

Quartile 3

Quartile 4

    
 

Non-FRL

FRL

Non-FRL

FRL

Non-FRL

FRL

Non-FRL

FRL

    

Quarterly earnings

$4,644.73

4,220.75

5,487.59

4,882.10

6,438.89

5,545.63

8,350.22

6,737.48

    
 

(3608.34)

(3262.46)

(4145.20)

(3664.24)

(4917.97)

(4359.39)

(6593.75)

(5459.76)

    

Educational attainment

            

 Voc/Cert

0.054

0.039

0.059

0.046

0.057

0.047

0.034

0.039

    

 AA

0.092

0.075

0.172

0.126

0.237

0.175

0.226

0.212

    

 BA

0.049

0.034

0.120

0.078

0.249

0.162

0.480

0.300

    

Enrollment

            

 Voc/Cert

0.144

0.144

0.122

0.113

0.102

0.095

0.057

0.073

    

 AA

0.552

0.536

0.610

0.556

0.614

0.559

0.479

0.521

    

 BA

0.100

0.082

0.209

0.161

0.376

0.284

0.601

0.460

    

Completion

            

 Voc/Cert

0.372

0.273

0.484

0.408

0.557

0.499

0.595

0.538

    

 AA

0.167

0.140

0.281

0.226

0.386

0.313

0.472

0.407

    

 BA

0.457

0.387

0.528

0.440

0.595

0.509

0.669

0.553

    

Intermediate HS outcomes

            

 10th grade FCAT math score

−1.189

−1.310

−0.184

−0.208

0.376

0.348

1.148

1.035

    
 

(0.748)

(0.813)

(0.17)

(0.171)

(0.166)

(0.163)

(0.413)

(0.333)

    

 HS GPA

2.032

1.941

2.395

2.283

2.730

2.579

3.201

2.973

    
 

(0.676)

(0.662)

(0.64)

(0.649)

(0.624)

(0.655)

(0.569)

(0.637)

    

Intermediate postsecondary outcomes

            

 Postsecondary GPA

1.826

1.657

2.174

2.022

2.461

2.317

2.865

2.682

    
 

(1.070)

(1.084)

(0.981)

(1.013)

(0.913)

(0.973)

(0.793)

(0.892)

    

 Total postsecondary credits

46.025

41.735

64.693

58.914

82.596

75.059

103.629

91.618

    
 

(49.98)

(48.566)

(56.496)

(56.072)

(59.317)

(60.512)

(60.924)

(62.933)

    

 Postsecondary math credits

5.686

5.225

8.102

7.714

8.885

8.961

9.461

9.694

    
 

(7.693)

(7.763)

(7.972)

(8.244)

(7.529)

(8.367)

(8.551)

(8.631)

    

 Postsecondary science credits

5.847

5.214

9.162

8.472

12.863

12.042

17.580

16.601

    
 

(11.001)

(10.675)

(13.896)

(14.353)

(17.937)

(18.004)

(22.205)

(22.442)

    

Labor market inputs

            

 Work experience

19.562

18.636

19.676

18.873

19.778

18.859

18.910

18.505

    
 

(11.556)

(11.36)

(11.118)

(10.987)

(10.767)

(10.743)

(9.978)

(10.317)

    

 Tenure

6.442

5.986

6.915

6.251

7.021

6.269

6.886

6.566

    
 

(7.188)

(6.715)

(7.387)

(6.775)

(7.177)

(6.604)

(6.687)

(6.603)

    

Notes: Quartile 1 is the lowest achievement level, and quartile 4 is the highest achievement level. Educational attainment and enrollment are conditional upon HS completion, and completion is conditional upon enrollment in that degree.

Table 3 presents the full range of higher education and labor market outcomes observed in earlier tables, but this time, they appear separately for individuals who fall into each of the four quartiles (where 1 is lowest and 4 is highest) on the high school math FCAT test. Parts A, B, and C of these tables reflect summary outcomes for all workers and students, then separately by race and FRL status respectively. In these tables, differences in outcomes observed between FCAT quartiles likely reflect the effects of selection by ability/achievement, while those differences within FCAT quartiles are likely attributable to other factors.

The results in Table 3A show large differences in both academic and labor market outcomes across FCAT quartiles. These differences are clearest and monotonic across achievement quartiles for earnings and tenure; BA attainment, enrollment and completion; and various measures of postsecondary achievement. In particular, BA enrollment and completion rates are heavily affected by selection on high school achievement. Enrollment in and completion of certificate and AA programs are less clearly affected by such selection.

Additionally, regardless of quartile, we find low completion rates in AA and BA programs. Specifically, conditional on enrollment, we find that 46% and 66% in the top quartile complete their AA and BA programs respectively; the comparable numbers in the second quartile are 26% and 50% respectively. These findings suggest that a range of other factors influence higher education outcomes as substantial numbers of even the highest achieving students fail to earn degrees.

These findings are generally confirmed in Tables 3B and C, which show the results broken out by race and FRL status. While differences in academic and labor market outcomes are partly accounted for by differences in high school achievement across these groups, important differences remain within particular levels of achievement. For instance, we find that black BA enrollments and postsecondary credits are often higher than those of whites within achievement quartiles, but their completion rates at all levels lag behind, as do their labor market earnings and attainment of work experience. In addition, FRL students lag behind non-FRL students in BA enrollments (and in AA enrollments below the top quartile), all intermediate postsecondary outcomes, and in earnings.

Within achievement quartiles, differences in educational attainment are large and quantitatively important, and we need to understand their determinants much better than we presently do.

3.3 Fields of study

It is well known that, for any level of academic attainment, labor market rewards differ significantly across fields of study (Altonji et al. 2012). In particular, the fields of science, technology, engineering and math (or STEM) are relatively highly rewarded.

In Table 4, we present the distributions of students across fields of study. We present these distributions separately for those in Vocational/Certificate, AA, and BA programs; and, within those, separately for completers and all attempters. If we compare ratios of concentration levels of completers to attempters by field, we can infer differential completion rates across these fields at each level of schooling. We present results for all students and workers, although results are similar by FRL status and race.
Table 4

Distribution of field of study by degree type and completion

 

Voc/Cert

AA

BA

Fields of study

Comp.

Attempt

Comp.

Attempt

Comp.

Attempt

Manufacturing

0.007

0.009

0.001

0.001

0.000

0.000

Construction

0.085

0.073

0.001

0.005

0.000

0.000

Health tech

0.258

0.130

0.024

0.023

0.000

0.001

Other health

0.100

0.066

0.055

0.072

0.078

0.068

Transportation

0.004

0.004

0.002

0.001

0.000

0.000

Business/Management

0.055

0.044

0.008

0.025

0.216

0.192

Education

0.013

0.025

0.003

0.007

0.085

0.080

Engineering

0.045

0.040

0.018

0.028

0.079

0.084

Communications

0.005

0.003

0.003

0.004

0.060

0.050

Legal

0.001

0.003

0.392

0.122

0.006

0.005

Security

0.332

0.198

0.007

0.016

0.042

0.038

Bio, Math/Stats, Physical science

0.005

0.004

0.002

0.003

0.069

0.078

Social science

0.000

0.001

0.001

0.003

0.203

0.180

Humanities

0.009

0.026

0.480

0.554

0.122

0.124

Other

0.078

0.053

0.003

0.006

0.034

0.030

Missing

0.002

0.322

0.001

0.131

0.006

0.072

N

12,060

28,775

43,236

144,231

45,564

77,028

Notes: “Comp.” denotes completers and “Attempt” denotes all attempters of that degree, regardless of completion status. Each number represents the share of students attempting a given degree who completed or attempted (depending on column header) a given field. For example, of students who completed an AA degree, 48 percent of them were awarded degrees in Humanities; and of students who attempted an AA degree, 55.4 percent attempted a Humanities degree. Students who did not complete a degree are counted as attempting their final declared major prior to leaving postsecondary studies.

Among our results, we find that vocational certificate students tend to concentrate heavily in Health Technology and Security, and to a lesser extent Construction and Other Health, and their completion rates are high in all these categories. Students at the AA level are concentrated in Legal studies and especially Humanities, but completion rates are relatively low in the latter. At the BA level, concentrations are high in Business/Management and the Social Sciences, and to a lesser extent in Other Health, Education, Engineering, Humanities, and Math/Physical Sciences; and completion rates are relatively low in Engineering, Math/Physical Sciences and the Humanities.

The low concentration in STEM fields by students in two-year institutions – who tend to come from disadvantaged backgrounds and/or with lower academic performance – may reflect rational decision-making, as STEM programs are associated with higher study times, harsher grading standards, and higher dropout rates for students with lower academic preparation (Arcidiacono et al. 2012; Arcidiacono et al. 2013). On the other hand, their relatively high concentration in other low-completion areas, like Humanities at the AA level (which is almost always “liberal arts” or “general studies” and designed to lead to transfer to four-year institution), is potentially more troubling given the low returns to these degrees, as shown in the next section.

3.4 Regression results

Tables 5, 6 and 7 show the results of estimated versions of Equation 1. The dependent variable is the natural log of quarterly earnings, and the sample is limited to those 18 years of age and older who have exited, or never entered, postsecondary education. The wage data are available for 5 years following a student’s last educational enrollment17.
Table 5

Regression results for log quarterly earnings: all workers

 

(1)

(2)

(3)

(4)

(5)

Highest credential

     

 No HS

−0.15***

−0.13***

−0.11***

−0.11***

−0.11***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

 Voc_Cert

0.31***

0.26***

0.29***

0.18***

0.18***

 

(0.01)

(0.01)

(0.01)

(0.01)

(0.01)

 AA

0.34***

0.36***

0.40***

0.10***

0.02**

 

(0.01)

(0.01)

(0.01)

(0.01)

(0.01)

 AS

    

0.29***

     

(0.02)

 BA

0.61***

0.69***

0.78***

0.39***

0.39***

 

(0.01)

(0.01)

(0.01)

(0.01)

(0.01)

 MA_PhD

0.89***

0.99***

1.08***

0.66***

0.66***

 

(0.01)

(0.01)

(0.01)

(0.02)

(0.02)

Work experience

0.02***

0.02***

0.02***

0.02***

0.02***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Work experience2

−0.00***

−0.00***

−0.00***

−0.00***

−0.00***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Tenure

0.08***

0.08***

0.08***

0.08***

0.08***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Tenure2

−0.00***

−0.00***

−0.00***

−0.00***

−0.00***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

FCAT

0.07***

0.06***

0.06***

0.04***

0.04***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Female

−0.15***

−0.15***

−0.15***

−0.16***

−0.16***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

FRL

−0.07***

−0.07***

−0.06***

−0.06***

−0.06***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Black

−0.11***

−0.12***

−0.12***

−0.11***

−0.11***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Hispanic

0.16***

0.15***

0.14***

0.15***

0.15***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Asian

0.08***

0.07***

0.07***

0.06***

0.06***

 

(0.01)

(0.01)

(0.01)

(0.01)

(0.01)

Other

−0.08***

−0.08***

−0.08***

−0.08***

−0.08***

 

(0.02)

(0.01)

(0.01)

(0.01)

(0.01)

CC Credits

     

 0 credits

  

0.20***

  
   

(0.01)

  

 1-15 credits

  

0.28***

  
   

(0.01)

  

 16-30 credits

  

0.28***

  
   

(0.01)

  

 31-45 credits

  

0.30***

  
   

(0.01)

  

 46-60 credits

  

0.32***

  
   

(0.01)

  

 60-80 credits

  

0.32***

  
   

(0.01)

  

University credits

     

 0 credits

  

0.02

  
   

(0.03)

  

 1-30 credits

  

0.26***

  
   

(0.01)

  

 31-60 credits

  

0.34***

  
   

(0.01)

  

 61-90 credits

  

0.40***

  
   

(0.02)

  

 91-120 credits

  

0.48***

  
   

(0.03)

  

 121-150 credits

  

0.55***

  
   

(0.03)

  

Post-secondary credits earned (100s)

     

 CC credits

 

0.57***

 

0.21***

0.20***

  

(0.01)

 

(0.01)

(0.01)

 University credits

 

0.47***

 

0.23***

0.24***

  

(0.01)

 

(0.01)

(0.01)

Post-secondary GPA

   

0.10***

0.10***

    

(0.00)

(0.00)

Post-secondary credits (10s)

     

 Math

   

0.04***

0.05***

    

(0.00)

(0.00)

 Science

   

0.02***

0.02***

    

(0.00)

(0.00)

Observations

3,739,354

3,739,354

3,739,354

3,739,354

3,739,354

R-squared

0.36

0.36

0.36

0.36

0.36

Notes: ***p < 0.01, **p < 0.05, *p < 0.1. The dependent variable is ln(quarterly earnings). Robust standard errors are in parentheses. FCAT scores are standardized to mean 1, standard deviation 0. Postsecondary GPA is measured on a 4 point scale. Quadratic and cubic FCAT terms are estimated but not reported.

Table 6

Regression results for log quarterly earnings by fields of study and level of education

 

AA & below

BA & above

Highest credential

  

 No HS

−0.12***

 
 

(0.01)

 

 Voc_Cert

−0.05

0.40***

 

(0.12)

(0.09)

 AA

0.19

0.31***

 

(0.12)

(0.07)

 BA

 

0.48***

  

(0.07)

 MA_PhD

 

0.73***

  

(0.07)

Completed field of study

  

 Manufacturing

0.26***

−1.87***

 Construction

0.25***

0.29*

 Health Tech

0.23***

0.16**

 Other Health

0.48***

0.39***

 Transportation

0.30**

0.81***

 Business/Management

0.08*

0.29***

 Education

−0.38***

0.03*

 Engineering

0.06

0.47***

 Communications

−0.21**

0.15***

 Legal

−0.01

0.10***

 Security

0.32***

0.07***

 Bio, Math/Stats, Phys Sci

0.15**

0.20***

 Social Science

−0.80***

0.07***

 Other

−0.08*

0.13***

 No CIP

0.02

0.31***

Work experience

0.02***

0.02***

 

(0.00)

(0.00)

Work experience2

−0.00***

−0.00***

 

(0.00)

(0.00)

Tenure

0.08***

0.05***

 

(0.00)

(0.00)

Tenure2

−0.00***

−0.00***

 

(0.00)

(0.00)

FCAT

0.06***

0.07***

 

(0.01)

(0.01)

Female

−0.13***

−0.08***

 

(0.01)

(0.01)

FRL

−0.05***

−0.02*

 

(0.01)

(0.01)

Black

−0.08***

−0.07***

 

(0.01)

(0.01)

Hispanic

0.15***

0.08***

 

(0.01)

(0.01)

Asian

0.03

−0.01

 

(0.02)

(0.02)

Other

−0.08**

−0.01

 

(0.03)

(0.05)

Observations

635,272

264,516

R-squared

0.24

0.19

Notes: ***p < 0.01, **p < 0.05, *p < 0.1. The dependent variable is ln(quarterly earnings). Robust standard errors are in parentheses. FCAT scores are standardized to mean 1, standard deviation 0. Quadratic and cubic FCAT terms are estimated but not reported.

Table 7

Regression results for log quarterly earnings by demographic and fcat groups

 

Gender

FRL

FCAT

 

Male

Female

No

Yes

Top half

Bottom half

Highest credential

      

 No HS

−0.12***

−0.19***

−0.14***

−0.16***

−0.11***

−0.15***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.01)

(0.00)

 Voc_Cert

0.35***

0.23***

0.29***

0.35***

0.24***

0.34***

 

(0.01)

(0.02)

(0.01)

(0.02)

(0.02)

(0.01)

 AA

0.29***

0.35***

0.31***

0.40***

0.25***

0.42***

 

(0.01)

(0.01)

(0.01)

(0.02)

(0.01)

(0.01)

 BA

0.64***

0.57***

0.59***

0.68***

0.51***

0.69***

 

(0.01)

(0.01)

(0.01)

(0.01)

(0.01)

(0.01)

 MA_PhD

1.03***

0.81***

0.86***

1.01***

0.80***

0.96***

 

(0.02)

(0.02)

(0.01)

(0.04)

(0.02)

(0.03)

Work experience

0.02***

0.02***

0.02***

0.02***

0.02***

0.02***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Work experience2

−0.00***

−0.00***

−0.00***

−0.00***

−0.00***

−0.00***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Tenure

0.08***

0.07***

0.07***

0.08***

0.06***

0.08***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

Tenure2

−0.00***

−0.00***

−0.00***

−0.00***

−0.00***

−0.00***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

FCAT

0.05***

0.09***

0.07***

0.07***

0.13***

0.07***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.02)

(0.02)

FRL

−0.06***

−0.09***

  

−0.06***

−0.07***

 

(0.00)

(0.00)

  

(0.01)

(0.00)

Female

  

−0.15***

−0.15***

−0.13***

−0.15***

   

(0.00)

(0.00)

(0.00)

(0.00)

Black

−0.17***

−0.03***

−0.13***

−0.08***

−0.07***

−0.12***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.01)

(0.00)

Hispanic

0.11***

0.21***

0.11***

0.20***

0.13***

0.16***

 

(0.00)

(0.00)

(0.00)

(0.00)

(0.01)

(0.00)

Asian

0.03**

0.13***

0.05***

0.13***

0.07***

0.07***

 

(0.01)

(0.02)

(0.01)

(0.02)

(0.02)

(0.01)

Other

−0.14***

−0.00

−0.10***

−0.05**

−0.03

−0.10***

 

(0.02)

(0.02)

(0.02)

(0.02)

(0.03)

(0.02)

Observations

2,096,183

1,656,973

2,196,457

1,556,699

1,052,977

2,700,179

R-squared

0.35

0.37

0.36

0.33

0.35

0.33

Notes: ***p < 0.01, **p < 0.05, *p < 0.1. The dependent variable is ln(quarterly earnings). Robust standard errors are in parentheses. FCAT scores are standardized to mean 1, standard deviation 0. Quadratic and cubic FCAT terms are estimated but not reported.

Five specifications of Equation 1 appear in Table 5. The first includes dummy variables for highest educational attainment, quarters of work experience and tenure with the current employer (in linear and quadratic form)18, a cubic polynomial of high school math FCAT score (to control for academic selection effects)19, and demographic dummy variables (race, gender, and FRL eligibility). The second equation adds the number of credits earned at either a two- or four-year school for students who enrolled but did not complete a degree. The third equation allows for non-linear returns to college credits by allowing the return to vary by the number of credits earned20. In the fourth equation, we add additional controls for postsecondary academic achievement: GPA and numbers of postsecondary courses taken in math and science. In the fifth equation, the dummy variable for AA degrees is split into those obtaining an Associate of Applied Science (AAS) or Associate of Science (AS) degree versus all others.

Several important findings appear in Table 5. As expected, we find strong labor market returns on average to all postsecondary credentials. Specifically, those obtaining a vocational certificate earn approximately 30 percent more per quarter than high school graduates; those earning AA degrees earn 35–40 percent more than high school graduates – though returns to AAS/AS degrees are much higher than those for other AAs (in Column 5); and the BA degree earns 60–80 percent more per quarter, which is about double what AA degrees earn, and those with graduate degrees earn considerably more21. Our estimates of returns to these credentials are at the high end of the range found in the other recent studies using state-specific administrative data and are a bit higher than (but still qualitatively consistent with) those estimated recently using national survey data.

The results in the second and third equations indicate that there are also returns, on average, to attending a program and earning credits, even if the program is not completed. In addition, there are larger returns (relative to those with no postsecondary enrollment) to accumulating more credits22. But these returns are smaller than those for completed degrees, indicating a “sheepskin effect,” especially for bachelor’s degrees. The magnitudes are broadly consistent with others who have found evidence of these effects in the literature (e.g., Kane and Rouse, Jepsen et al., Kreisman et al., op. cit.)23. Furthermore, since most dropouts accumulate many fewer credits than program completers, the dropouts are relatively hurt both by their fewer credits and by the absence of the formal credential. Thus, the low levels of completion observed in AA programs, and the gaps between race, income, and achievement groups at the BA level all reduce the future earnings of minority and low-income students.

There is also a very strong return to postsecondary GPA and also a return to taking math and science courses. When these variables are added to the equation, the estimated returns to completed degrees drop substantially, which indicates that at least some part of the observed returns to degree programs are really for overall achievement and technical skills24. On the other hand, it could also be the case that high postsecondary GPA captures unmeasured ability in a way our other variables do not, thus explaining additional selection into degree program. The relatively high returns to mostly technical certificate programs, and to AAS/AS degrees rather than AAs, confirm that the market returns to technical skills, including at the sub-BA level, are relatively large 25.

A few other findings in Table 5 are noteworthy. First, the returns to general work experience are modest, but in contrast, the returns to tenure are quite substantial, with a first quarter return of about eight percent. Thus, accumulating job tenure for a year or more is quite substantially rewarded. Though we made no effort here to measure returns to work experience specifically tied to one’s postsecondary degree – which presumably are much greater than these – the results indicate that accumulating some kinds of work experience before, during or after one’s postsecondary schooling can be valuable. In addition, the returns to FCAT indicate some significant academic selection effects, which we control for in our analysis26.

Finally, we note the coefficients on demographic variables in these equations. Interestingly, controlling for the full range of academic attainment and achievement generates even larger negative effects for females – of 15 percent per quarter – than we found in Table 1, since their achievement measures are generally higher than those of males, while their earnings are lower. Controls for education and achievement also generate large (15 percent) returns for Hispanics relative to whites. And, for blacks and low-income students, estimated negative effects are now much smaller (6 to 12 percent per quarter) than we observed in Table 1, indicating that much (though not all) of their lower earnings is associated with lower academic attainment and achievement27.

Despite our extensive controls for demographic information and achievement in high school, it is still possible that our estimates reflect selection on unobserved factors. For instance, for two students with the same demographic information and performance in high school who choose two different postsecondary paths and have different labor market outcomes, we cannot say for certain whether it was the paths that caused the disparate outcomes or some unobserved factor such as motivation or guidance from others. On the other hand, including controls for high school achievement generates only modest changes in our estimated coefficients on schooling, relative to equations which included no such controls. Thus, it seems unlikely that other unobservable characteristics have substantial effects on these estimates28.

3.5 Specific fields, specific groups

To further explore how these average effects on earnings vary with field of study, we present returns to different fields in Table 6. These estimates are from versions of Equation 1 containing dummy variables for fields of study, with Humanities as the omitted group. In separately estimated equations, we present results for those earning AAs or certificates and those earnings BAs.

The results show substantial variation in returns across fields of study. In particular, those earning sub-BA credentials have relatively strong returns to health, transportation, construction, manufacturing (mostly certificates), and security credentials; and for those earning BAs or higher, returns are strongest in transportation, engineering, business management and health, with smaller but still positive pay premia (relative to humanities) in the math/science, communications, legal work, and health technology29.

Importantly, the returns to the omitted category of AAs and humanities – the most commonly pursued degree in two-year colleges and also a field with relatively low completion rates – are also relatively low, compared to virtually all other fields. A strong case can be made therefore that the large concentrations of AA students in these low completion and low return areas are sub-optimal, at least in terms of future earnings potential.

Our final research question is how these market returns vary across different demographic groups of students or those with varying levels of earlier achievement. To answer this question, we estimate the model (also based on the first equation in Table 5) separately for males and females, those eligible or not eligible for FRL, and those in the top or bottom half of FCAT scores30. These estimates appear in Table 7.

Overall, the patterns of market returns to postsecondary attainment and achievement are quite similar across demographic and achievement groups. A few modest differences can be found. For instance, men earn relatively larger rewards than women, both in certificate programs and BAs and above, while women do better in AA programs. In addition, academic credentials, including vocational certificates and AAs, are also relatively well-rewarded for FRL students and those students in the bottom half of high school achievers.

Thus, labor market rewards are broadly similar across these groups. But, in some more technical fields, achieving a vocational certificate or even an AA can be quite rewarding, especially among lower-achieving groups, and particularly if they can combine these credentials with good work experience or tenure afterwards.

4 Conclusion

Our paper examines a range of issues focusing on postsecondary education and labor market outcomes, with a particular focus on minorities and/or disadvantaged workers. We use administrative data from the state of Florida, where secondary and postsecondary education records for two cohorts of students have been linked to UI earnings data (for five years after schooling has been completed). The administrative data give us enormous samples with which to study particular groups of students, with very detailed information on educational experiences and outcomes for every public school student in a particular year. In addition, the secondary education records to which we have access enable us to test and control for selection based on early academic experience and achievement to a much greater extent than has been the case with other studies using administrative data. However, the data have their limitations, for instance, we cannot follow those who leave the state of Florida to attend college or enter the labor force. Yet, the richness of the data on educational experiences and outcomes, coupled with the recentness of the cohort and our large sample size, enable us to learn a great deal about postsecondary outcomes and their determinants.

Our main findings can be summarized as follows:
  1. 1)

    Variation in secondary school achievement can account for a large part of the gap in postsecondary attainment (through the differences we observe in enrollment rates and especially in completion) and labor market outcomes between racial and family income groups, but quite large differences also exist within achievement groups;

     
  2. 2)

    Earnings of minorities and disadvantaged students are reduced not only by their lower educational attainment and completion but also by weaker postsecondary academic performance and by their chosen fields of study.

     

More specifically, significant labor market premia are available in a variety of more technical certificate and AA programs, even for those with weak earlier academic performance. However, many disadvantaged (and other) students end up in general humanities (or liberal studies) programs at the AA (and even the BA) level, with low completion rates and low compensation afterwards.

Of course, students in liberal arts curricula, and especially those planning on post-BA education, are not necessarily harmed by majoring in humanities. And in none of this work can we control for students’ preferences across fields; therefore, we cannot infer whether these outcomes reflect sub-optimal choices on the part of students.

On the other hand, it is likely that many students in AA programs, especially disadvantaged students or those with weaker academic performance, do not plan on obtaining post-BA education and are hoping that their college experiences lead directly to higher earnings. For these students, the very high concentrations we observe in humanities programs at the AA level do not appear consistent with those goals.

One can imagine a variety of reallocations of students from these programs that would improve expected student outcomes, especially among the disadvantaged. These reallocations might include moving the higher achievers in this group to BA programs (especially at more selective institutions where completion rates are relatively high) or to more technical AA fields of study with higher completion rates and earnings; while lower achievers might do better in some of the well-compensated vocational certificate programs. Additionally, certificate programs and work experience are particularly well-compensated among young men, especially African-American men who have difficulty gaining work experience and whose postsecondary attainments also lag substantially behind those of women.

How might such reallocations be accomplished? Poor choices by postsecondary students likely reflect at least two problems: poor information among students and poor incentives faced by their postsecondary institutions. Due to an unstructured environment and poor counseling, students in community college receive very little information about either academic or job market opportunities. Improving the guidance provided to these students would likely improve their outcomes (Scott-Clayton 2011; Jenkins and Cho 2012; Rosenbaum et al. 2006; Jacobson and Mokher 2009). But public institutions also need more incentive to respond to labor market factors and prepare students for well-paying or higher-demand fields, even if the costs to them of instruction and equipment in these fields are higher (Holzer 2014). Incentivizing these institutions to build more partnerships with industry associations, generating sectorial and career pathway programs, and helping students participate in them could be accomplished by greater use of postsecondary and earnings performance measures in determining state subsidies for public colleges and universities31.

In addition to these approaches, a few other broad policy guidelines are at least consistent with, and are perhaps suggested by, our findings. In particular, disadvantaged students would benefit from reforms in support programs and services, such as developmental (or remedial) education and financial aid32. Students might also benefit from work-based learning models and other pathways to postsecondary education, such as apprenticeships and other forms of high-quality career education, which do not substitute work experience for postsecondary learning but complement it and lead to improved labor market outcomes33. Experimentation with, and further evaluation of reforms in these support services and alternative pathways to skill creation should be a high priority.

4.1 Endnotes

1See Jacobson and Mokher (2009), Jepsen et al. (2014) and Kreisman et al. (2013).

2The unemployment insurance records do not include information on several small categories of employees including self-employed and federal workers; but they cover roughly 96% of all workers (Andersson et al. 2005).

3Institute of Education Sciences (2014). Digest of Education Statistics, Table 232. Retrieved May 21, 2014, from http://nces.ed.gov/programs/digest/d11/tables/dt11_232.asp

4Institute of Education Sciences. Digest of Education Statistics, Table 233. Retrieved May 21, 2014, from http://nces.ed.gov/programs/digest/d12/tables/dt12_233.asp.

5Malamud and Wozniak (2008).

6For instance, in our oldest cohort, over 70% of those in the top achievement quartile appear in the labor market, whereas for the lowest quartile the comparable fraction is just 35%. We examine the robustness of our results to including those with zero earnings for very limited numbers of quarters (see footnote 9), but we cannot address those with longer periods of no earnings (who might have left the state or left the labor market for a variety of reasons). The omission of these nonparticipants from our findings might bias downwards our estimates of returns to postsecondary credentials since the nonparticipants are likely high school graduates and dropouts who would have had low earnings had they worked. The same downward bias likely results from the omission of higher achievers and earners who moved out of state.

7In national data, the percentages of enrollees at community and public four-year college who are above the age of 24 are 40 and 21 percent, respectively (National Center for Education Statistics 2013, Table 303.50 of Digest 2013).

8Our measures of work experience and tenure do not count intermittent quarters with zero employment towards those measures.

9Researchers generally attribute quarters with zero earnings to labor supply decisions in which the individual did not choose to join the labor force or could not find employment; however, in our data, another possible explanation is that the individual moved out of state. To assess whether including short-term non-working spells are driving our results, we conduct a robustness check in which we impute earnings for individuals with 1–3 quarters of missing UI records in between observed employment spells. The results are robust to this imputation of zero earnings.

10While it is still possible that differences in outcomes within quartiles could reflect differences in achievement given the broadness of the quartiles, we use a variety of specifications for achievement measures in our earnings regressions below, including decile dummies, and find quite similar effects.

11In our descriptive tables, we pool whites and Asians, though in our regression estimates below we separate them (whites are the omitted group and Asians are indicated by a dummy variable). Though Asians earn more than whites, even controlling for education and achievement, their numbers are too small to generate major inconsistencies between the earlier descriptive results and our regression estimates.

12Regression results are qualitatively very similar when controlling for high school GPA and FCAT reading scores in addition to FCAT math scores. However, high school GPA is missing for one of our cohorts, so we choose not to include it (the correlation between FCAT math and high school GPA is just above 0.7). Since adding reading FCAT scores generates virtually identical results when already controlling for math FCAT scores, we focus only on math scores for simplicity.

13We present quarterly, rather than annualized, earnings to be consistent with previous studies (e.g., Jepsen et al. 2014). To eliminate outliers, we limited our sample of quarterly earnings between $100 and $100,000.

14Swanson (2004) presents high school dropout rates using administrative data, though Mishel and Roy (2004) compare them to survey-based estimates, which are much higher. Using either method, Murnane (2013) shows a large decline in the estimated dropout rate over the past decade.

15When examining Hispanics separately by subgroup in the subsequent earnings regressions, relative to whites, all Hispanic groups earned more on average. The highest Hispanic earners are Cuban-born (about $1,200 extra per quarter) and lowest earners are Puerto Rican-born ($170 extra per quarter).

16While surveys tend to undercount low-income men, especially those who have been previously incarcerated or who are only marginally attached to households, the administrative data will undercount men or women whose employment is sporadic and informal, especially those paid in cash.

17By focusing on students’ earnings after their education has been completed, we avoid confounding those with periods of lower earnings before or during the attainment of postsecondary credentials.

18Job tenure and experience are likely endogenous as they are partially driven by postsecondary choices and outcomes. Results are qualitatively similar when not controlling for tenure and/or experience, although not controlling for either somewhat depresses the coefficient on BA completion since BA recipients in each cohort tend to have had less time in the labor market and thus lower earnings (conditional on degree earned) relative to others of the same age who have accrued more labor market experience.

19Variables with missing values for all of these measures are used in the regressions, with zeroes imputed for the missing values and dummy variables used to denote such cases. Results are robust to using a linear term in math FCAT scores or including a dummy variable for each decile of math FCAT achievement. As the coefficients on the quadratic and cubic FCAT terms are small, we only report the coefficient on the linear term in regression results.

20Many community colleges in Florida offer BA degrees; however, the nature of the data makes it difficult to distinguish credits earned at a community college in pursuit of a BA from credits earned at a community college in pursuit of a different degree. To ensure a clean measure of credits earned in pursuit of a BA (four-year credits), we drop the approximately 500 students who attempted a BA at a two-year institution and the 100 students who earned a BA at a two-year institution.

21Kane and Rouse (1998) report that a year of schooling at the two-year level is valued similarly in the labor market to one from a four-year school. But Acemoglu and Autor (2011) report a recent “convexification” of the returns to schooling in which each year of additional postsecondary schooling generates higher average value for all.

22The relatively large return to CC students with zero credits is puzzling. It might indicate some selection into these schools by unobservables not captured by our FCAT scores or other variables. In contrast, the near-zero return to those with no credits in BA programs suggests no such selection exists for these students.

23An exception to this literature is Clark and Martorell (2014) who find little evidence of sheepskin effects for high-school diplomas.

24When postsecondary GPA and math/science credits are added separately, the former generates a much larger decline in the value of the degree than the latter. Details are available from the authors. Also, the coefficients on math/science credits in Col. 4 cannot be directly compared to the returns to all credits since the latter is only for degree non-completers, while the former is for all students.

25See the Economic and Statistics Administration (2011) for recent evidence on the labor market values of STEM jobs, and Jacobson et al. (2005) for earlier evidence on the values of community college education with more technical courses and curricula, though for a sample of older displaced workers.

26Other attempts to control for selection (such as with high school GPA) generate similar results. Since FCAT has fewer missing observations, we include only FCAT scores.

27Johnson and Neal (1998) and Holzer and Dunlop (op. cit) show that racial gaps in earnings grow much smaller when we control for differences in education plus academic achievement. But the inclusion of quarters with zero earnings can strongly reduce the extent to which education and achievement account for the earnings gaps of black, relative to white, men.

28In estimated equations that contain no controls for high school achievement, our coefficients on having AA, BA, or graduate degrees in column 1 are .03, .04, or .07 higher, respectively, and the broad qualitative pattern of returns remains the same.

29We find similar patterns across fields of study among those not completing degree programs, although the results are not quite as pronounced. Results are available from the authors.

30We have estimated separate equations by race; findings are similar to the FRL results and are available from the authors.

31The National Governors’ Association (2013) reports a widespread attempt to build “sectorial” training partnerships at the state and regional levels between community colleges and industry groups, though we have little data on the scale of student enrollments in programs generated through these partnerships. The National Council of State Legislatures (2014) reports that over half of all states are planning to at least partially use higher education academic outcomes to determine state subsidies to higher education institutions, and Holzer (2014) argues that employment outcomes should also be used to measure institutional performance as well. But the use of such outcomes without adjusting for the quality of student inputs could result in “cream-skimming” through higher admissions standards, among other potential unanticipated consequences.

32Long (2014) notes that remedial classes, to which many academically under-performing students are assigned in community colleges before they are allowed to take courses for credit, are generally ineffective and recommends a range of reforms. The College Board (2013) report on Pell grants also recommends a range of reforms designed to improve completion rates and subsequent earnings among low-income students.

33Models of high-quality career and technical education, such as apprenticeship, increasingly build the attainment of postsecondary credentials like AA degrees into their training model. See Lerman (2010).

Declarations

Acknowledgements

This paper is part of the Postsecondary Education and Labor Market Program at the Center for the Analysis of Longitudinal Data in Education Research (CALDER) at AIR. We thank Kennan Cepa and Tiffany Chu for excellent research assistance. We also thank Ken Troske and other participants at the Conference on Building Human Capital, Institute for Research on Poverty, University of Wisconsin, July 2014; the CAPSEE conference, Washington DC, September 2014; and an anonymous referee for their discussion and comments. This research was supported by the CALDER postsecondary initiative, funded through grants provided by the Bill and Melinda Gates Foundation and the Smith Richardson Foundation to the American Institutes of Research.

Responsible editor: David Neumark

Authors’ Affiliations

(1)
American Institutes for Research (AIR)/CALDER
(2)
Georgetown University and AIR/CALDER
(3)
RTI International

References

  1. Acemoglu D, Autor D (2011) Skills, Tasks and Technologies: Implications for Employment and Earnings. In: Ashenfelter O, Card D (eds) The Handbook of Labor Economics, vol 4, North Holland, AmsterdamGoogle Scholar
  2. Altonji, J. Blom E, Meghir C 2012. “Heterogeneity in Human Capital Investments: High School Curriculum, College Major, and Careers,” National Bureau of Economic Research Working Paper, Cambridge MA
  3. Andersson F, Holzer H, Lane J (2005) Moving Up or Moving On: Who Advances in the Low-Wage Labor Market? Russell Sage Foundation, NewYorkGoogle Scholar
  4. Arcidiacono P, Aucejo E, Spenner K (2012) “What happens after enrollment? an analysis of the time path of racial differences in GPA and major choice,” IZA J Labor Econ 1:Article 5View ArticleGoogle Scholar
  5. Arcidiacono P, Aucejo E, Hotz VJ (2013) “University Differences in the Graduation of Minorities in STEM Fields: Evidence from California”, revise and resubmit American Economic Review View ArticleGoogle Scholar
  6. Bahr P, Dynarski S, Jacob B, Kreisman D (2014) “Labor Market Returns to Community College in Michigan.” Unpublished manuscript, University of MichiganGoogle Scholar
  7. Bailey M, Dynarski S (2011) Inequality in Postsecondary Education. In: Duncan G, Murnane R (eds) Whither Opportunity? Rising Inequality, Schools, and Children’s Life Chances. Russell Sage Foundation, New YorkGoogle Scholar
  8. Belfield C, Liu YT, Trimble MJ (2014) “The Medium-Term Labor Market Returns to Community College Awards: Evidence From North Carolina, New York. Center for Analysis of Postsecondary Education and Employment, NYGoogle Scholar
  9. Borjas G (1987) Self-selection and the earnings of immigrants. Am Econ Rev 77(4):531–553Google Scholar
  10. Bound J, Lovenheim M, Turner S (2009) Why Have College Completion Rates Declined? An Analysis of Changes in Student Preparation and College Resources. National Bureau of Economic Research Working Paper, Cambridge MAView ArticleGoogle Scholar
  11. Clark, D and P Martorell. 2014. “The Signaling Value of a High School Diploma” (with Paco Martorell). Journal of Political Economy. Forthcoming.
  12. College Board. 2013. Rethinking Pell Grants. College Board, Washington DC
  13. Economic and Statistics Administration (2011) STEM: Good Jobs Now and For the Future. Department of Commerce, USGoogle Scholar
  14. Haskins R, Holzer H, Lerman R (2009) Promoting Economic Mobility by Increasing Postsecondary Education. The Pew Trusts, Washington DCGoogle Scholar
  15. Holzer H (2014) Improving Employment Outcomes for Disadvantaged Students. Hamilton Project, The Brookings Institution, Washington DCGoogle Scholar
  16. Holzer H, Dunlop E (2013) Just the Facts, Ma’am: Postsecondary Education and Labor Market Outcomes in the U.S. CALDER Working Paper, American Institutes for Research, Washington DCGoogle Scholar
  17. Hoxby C, Turner S (2013) Informing Students About their College Prospects: A Proposal for Broadening the Expanding College Opportunity Project. The Hamilton Project, Brookings Institution, Washington DCGoogle Scholar
  18. Institute of Education Sciences (2014) Digest of Education Statistics. U.S. Government Printing Office, Washington DCGoogle Scholar
  19. Isaacs, J (2007). Economic Mobility of Families across Generations. Policy Brief, Center on Children and Families, the Brookings Institution, Washington DC
  20. Jacobson L, Mokher C (2009) Pathways to Boosting the Earnings of Low-Income Students by Increasing their Educational Attainment. Hudson Institute, New YorkGoogle Scholar
  21. Jacobson L et al (2005) The Impact of Community College on Older Workers: Should We Teach Old Dogs New Tricks?”. Working Paper, Cornell ILR SchoolGoogle Scholar
  22. Jenkins D, Cho S-W (2012) Get with the Program: Accelerating Community College Entry Into and Completion of Programs of Study. Community College Research Center Working Paper, Teachers College, Columbia UniversityGoogle Scholar
  23. Jepsen C, Troske K, Coomes P (2014) The labor-market returns to community college degrees, diplomas, and certificates. J Labor Econ 32(1):95–121, University of Chicago PressView ArticleGoogle Scholar
  24. Johnson W, Neal D (1998) Basic Skills and the Black-White Earnings Gap. In: Jencks C, Phillips M (eds) The Black-White Test Score Gap. Brookings Institution, Washington DCGoogle Scholar
  25. Kane T, Rouse C (1998) The labor market returns to Two-and four-year colleges. Am Econ Rev 85(2):600–614Google Scholar
  26. Kreisman D, Dynarski S, Jacobs B, Bahr P (2013) The Returns to Couse Credits, Certificates and Degrees: Evidence from Michigan’s Community Colleges. Unpublished, University of MichiganGoogle Scholar
  27. Lerman R (2010) Expanding Apprenticeships: A Way To Improve Skills and Careers. The Urban Institute, Washington DCGoogle Scholar
  28. Long BT (2014) Addressing the Academic Barriers to Success. In: Kearney M, Harris B (eds) Policies to Address Poverty in America. The Hamilton Project, Brookings Institution, Washington DCGoogle Scholar
  29. Lovenheim M, Reynolds C (2011) Changes in postsecondary choices by ability and income: evidence from the national longitudinal surveys of youth. J Human Capital 5(1):70–109View ArticleGoogle Scholar
  30. Malamud O, Wozniak A (2008) The Impact of College Graduation on Geographic Mobility: Identifying Education Using Multiple Components of Vietnam Draft Risk, IZA Discussion Paper Series No. 3432Google Scholar
  31. Mishel L, Roy J (2004) Rethinking High School Graduation Rates and Trends. Economic Policy Institute, Washington DCGoogle Scholar
  32. Murnane R (2013) U.S. High school graduation rates: patterns and explanations. J Econ Lit 51(2):370–422View ArticleGoogle Scholar
  33. National Center for Education Statistics (2013) Digest of Education Statistics. Institute of Education Sciences, Washington DCGoogle Scholar
  34. National Council of State Legislatures. 2014. “Performance-Based Subsidies for Higher Education.” http://www.ncsl.org/research/education/performance-funding.aspx
  35. National Governors’ Association (2013) State Sector Strategies Coming of Age. Center for Best Practices, Washington DCGoogle Scholar
  36. Reardon S (2011) The Widening Academic Achievement Gap Between the Rich and the Poor: New Evidence and Possible Explanations.”. In: Duncan G, Murnane R (eds) Whither Opportunity? Rising Inequality, Schools, and Children’s Life Chances. Russell Sage Foundation, New YorkGoogle Scholar
  37. Rosenbaum J, Deil-Amen R, Person A (2006) After Admission: From College Access to College Success. Russell Sage Foundation, New YorkGoogle Scholar
  38. Scott-Clayton J (2011) The Shapeless River: Does a Lack of Structure Inhibit Students’ Progress at Community Colleges?” Community College Research Center Working Paper, Teachers College, Columbia UniversityGoogle Scholar
  39. Swanson C (2004) Who Graduates? Who Doesn’t? A Statistical Portrait of Public High School Graduation, Class of 2001. The Urban Institute, Washington, DCGoogle Scholar
  40. Xu D, Jaggers S (2014) Examining Wage Trajectories of Community College Students Using a Growth Curve Modeling ApproachGoogle Scholar

Copyright

© Backes et al.; licensee Springer. 2015

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.