Research in Higher Education

, Volume 49, Issue 3, pp 199–213

The Way to Wealth and the Way to Leisure: The Impact of College Education on Graduates’ Earnings and Hours of Work

Authors

    • Department of Leadership, Policy, and OrganizationsVanderbilt University
Article

DOI: 10.1007/s11162-007-9080-5

Cite this article as:
Zhang, L. Res High Educ (2008) 49: 199. doi:10.1007/s11162-007-9080-5

Abstract

This study extends the analysis of the economic return of college education up to 10 years after college education and further examines the impact of college education on graduates’ hours of work. The results suggest that variation in hours of work explains a portion of earnings differentials among college graduates. Graduates from high-quality private institutions tend to work longer hours than their peers from other types of institutions. Female graduates spend fewer hours working than their male counterparts. As far as family background is concerned, graduates from high-income families tend to work longer hours and first-generation college graduates tend to work fewer hours. Finally, business majors seem to work longer hours while health and public affair majors less hours.

Keywords

EarningsHours of workUndergraduate majorsCollege quality

Introduction

Recent studies on the impact of college education have focused on the earnings differentials among college graduates with different educational experiences (e.g., Behrman et al. 1996; Brewer and Ehrenberg 1996; Brewer et al. 1999; Dale and Krueger 2002; Thomas 2003; Thomas and Zhang 2005). Most studies along this line of inquiry are rooted in the tradition of human capital theory, which posits that the labor market rewards investments individuals make in themselves and that these investments lead to greater productivity and to higher salaries (Becker 1964). For example, one of the central features of econometric work in this area is the economic return of college quality. From a policy perspective, this question is important because students and their families often make great financial commitment to attend prestigious institutions. Such a commitment is based on the conventional wisdom that investing in a better college education will eventually lead to greater economic benefits. Empirical evidence in this area has generally confirmed that it pays to attend a prestigious institution (e.g., Brewer and Ehrenberg 1996; Brewer et al. 1999; Zhang 2005).

The economic analysis of college education makes perfect sense if individuals only consider the financial well-being of their lives. Nonetheless, it is not uncommon that individuals make trade-offs, most notably between their work and family (or leisure in general). For example, one might prefer a less demanding job (e.g., fewer hours of work) albeit with less earnings. On a theoretical level, because both earnings and leisure (non-working time) benefits individuals, it is important to add the latter to the theoretical framework that guides individuals to make important choices in college, such as college major and prestige.

The human capital framework seems less straightforward in explaining trade-offs individuals make between earnings and leisure unless one is able to convert the latter into amount of dollars. The principle of utility maximization might provide some guidance in this regard. Suppose an individual’s utility is an increasing function of both earnings and leisure time, it is straightforward that he or she prefers higher earnings and fewer hours of work, ceteris paribus. This framework could be useful in some simple scenarios even with the actual utility functional form unknown. For example, if two jobs provide same earnings, the one that demands fewer hours of work is preferred according to the utility maximizing principle. Admittedly, real life decisions are much more complicated, usually involving trade-offs between earnings and leisure time. For example, how to compare a job that provides 5% more earnings but also demands 3% more hours of work with another job that provides fewer earnings but more leisure time? The answer most likely varies by individuals, as determined by their idiosyncratic preferences. Nonetheless, it is important to examine how college choices relate to earnings and leisure because both provide valuable information for students to make sound decisions. For example, empirical evidence has already shown that it pays off economically to attend a high-quality institution. This conclusion could be enhanced if attending a high-quality institution is also found out to provide more leisure time to individuals. In contrast, if higher levels of earnings come with a sacrifice of leisure time, one might carefully consider both elements in college choice decisions.

The purpose of this study is to examine simultaneously the impact of college education on earnings and on leisure. This inquiry thereby extends the most recent research on earnings determination structure by considering another important aspect of work life. To achieve this objective, I pose the following questions: (a) What is the impact of various educational and non-educational factors on earnings for recent college graduates? (b) How are these educational and non-educational factors related to leisure? (c) And finally, which is the main mechanism through which college education influences the careers of college graduates? As a by-product of using the third follow-up survey of the Baccalaureate and Beyond (B&B) study, the current analysis also extends the research on the economic return of college education up to 10 years after college education.

Literature Review

This study is informed by earlier work of Thomas and Zhang (2005) on the impact of various educational and non-educational factors on earnings of recent college graduates. Results of their study and those of earlier researchers suggest that the earnings of college graduates are affected by an array of post-secondary educational experiences. Among them are academic majors, institutional selectivity, and academic performance. To summarize these studies, all other factors being equal, students graduating in higher-demand majors (Berger 1988; Eide 1994; Grogger and Eide 1995; James et al. 1989; Rumberger and Thomas 1993) from higher quality institutions (Brewer et al. 1999; Fox 1993; James and Alsalam 1993; Mueller 1988; Smart 1988; Rumberger and Thomas 1993) with better academic performance (James et al. 1989; Jones and Jackson 1990) tend to command higher salaries than their peers from lower-quality colleges and/or alternate academic majors.

Non-academic factors also greatly influence graduates’ earnings. Zhang (2005) suggests that this influence is mainly indirectly through educational attainment. For example, socioeconomic factors such as family income and parental education have a significant impact on the probability that a student will earn a degree at a highly selective institution. This influence may be exercised directly and in the short term (i.e., students of high-SES background are in a better position to pay for college education at high-cost institutions) or indirectly and in the long term (i.e., they are better prepared in terms of measured intellectual ability). This study shows that academic and non-academic factors tend to work in tandem.

While earnings were treated as one variable in those previous studies, labor economists have studied separately the two components of earnings: the wage rate and hours of work. For a given job (usually with predetermined wage rate), an individual has to decide whether he or she will work; and if one decides to work, then how many hours of work. Because individuals typically value their leisure time, they prefer not to work if the wage rate is below a certain threshold (i.e., reservation wage). If the given wage rate is above the individual’s reservation wage, he or she will participate in the labor force, and hours of work are consequently determined by how much he or she values the leisure time and income. In the language of labor economics, individuals will work to the point where the marginal rate of substitution between leisure and income equals the wage rate. That is, individuals derive the same level of utility either from spending the last hour at home or from the income earned by spending the last hour working.

According to this model of labor supply, hours of work are determined by the wage rate and other factors such as non-wage income and an individual’s tastes in leisure. Typically, when the wage rises within a certain range, individuals will increase their hours of work. If the wage rate continues to rise beyond a certain point, individuals may want to cut back their hours of work because the value of additional income starts to wane.1 The empirical evidence of the positive relationship between wage rate and hours of work is generally weak (Ehrenberg and Smith 2006), probably due to the so called “division bias” introduced by the measure error in hours of work (Borjas 1980). However, empirical evidence has shown that education has positive impact on both wage rate and hours of work. For example, Card (1999) used CPS data to examine the relationship between education and earnings and hours of work; he concluded that education in terms of schooling years had a positive impact on both the hourly wage rate and hours of work. An additional year of education is associated with an approximately 10% increase in hourly income and 5% increase in hours worked annually, yielding about 15% increase in earnings. These results suggest that better educated individuals enjoy higher earnings for two reasons: their hourly incomes are higher and their work hours are longer. That is, those who enjoy higher hourly income have more incentives to work longer hours to capitalize their investment in human capital. The current study extends this line of research by examining the impact of finer differences in educational attainment such as college quality and academic majors on the hourly income and hours of work.

One might wonder how a simple, static model of labor supply describes the employment decision of college graduates in the labor market. For example, a full-time job offer typically announces the annual salary but not the wage rate. Furthermore, hours of work are tightly keyed to the nature of job, and individuals do not have much discretion in deciding how many hours they would like to work. As such, one might want to relax the assumption that wage rate is predetermined and individuals choose how many hours to work. Instead, the hourly income could be viewed as the realized income per hour, determined by the salary and hours of work, both of the latter two are keyed to the nature of work. In the current analysis, because most college graduates are salaried workers (e.g., only 12.2% of full-time workers reported their earnings in hourly wage rate in 1997), it is probably more appropriate to examine how characteristics of college education are related to the earnings and hours of work experienced by college graduates.2 Based on these analyses, one could further derive the impact of college education on actual realized hourly income. (The calculation will be discussed in next section).

Examining the impact of college education on earnings and hours of work separately yields new insight into the mechanisms through which college education influence graduates’ life. Take a simple example, Thomas and Zhang (2005) estimate that graduating from a high-quality institution provides an earnings premium that is approximately 20% higher than graduating from low-quality colleges. This 20% earnings premium could result from different combinations of hourly income and hours of work. For example, this earnings premium could be solely due to an increased number of work hours for those graduates from high-quality institutions (i.e., college graduates from high-quality institutions could work 20% more hours than those from low-quality institutions), with the same hourly income for everyone. In contrast, if graduates from both types of colleges work the same number of hours, then the hourly income for those from high-quality institutions is 20% higher than those from low-quality institutions. In fact, any linear combination of the above two scenarios are all possible, including the combination that those from high-quality institutions work fewer hours than those from other institutions but the hourly income for the former is more than 20% higher than the hourly income for the latter.

It is necessary, then, to examine how various educational and non-educational factors affect graduates’ earnings and hours of work separately. Although quite a few studies have studied the earnings determination structure for college graduates, the impact of various educational and non-educational factors on hours of work is less obvious. Rough predictions are possible based on previous research (Card 1999) and anecdotal evidences. For example, Zhang and Thomas (2003) explore the impact of college education on graduates’ job satisfaction and find that graduating from high-quality institutions, although yielding significant earnings advantages over graduating from lesser institutions, could have a negative impact on graduates’ job satisfaction once the positive impact of earnings on job satisfaction is isolated. This lower job satisfaction could be due to longer hours of work or other less favorable working conditions for those college graduates from high-quality institutions. Another set of variables that might affect hours of work is undergraduate majors. To the extent that undergraduate majors significantly affect the kind of jobs that graduates would be offered and the variation in the number of working hours required across occupations/jobs, undergraduate majors could also influence the number of working hours.

Data and Methods

Data Set

The major data set used in this analysis is B&B: 93/97/03 provided by National Center for Education Statistics (NCES). B&B is a national longitudinal study designed to provide information concerning education and work experiences after completion of the baccalaureate degree. It provides (a) cross-sectional information 1 year after bachelor’s degree completion and (b) longitudinal data concerning entry into and progress through graduate-level education and the work force. The B&B study sampled more than 10,000 baccalaureate recipients who completed their degrees between July 1992 and June 1993. The first, second, and third follow-up surveys, conducted in 1994, 1997, and 2003 respectively, provide information on these college graduates’ experience during their early careers. Because prior research has shown that such institutional characteristics as selectivity are important influences on graduates’ earnings, I use the restricted-use data set to match students with institutions.3

Variables

Earnings and hours of work are the two main variables of interest in this study. Earnings are measured as the annualized self-reported earnings at each of the three follow-up survey points. Hours of work is measured by the self-reported number of working hours per week. For example, the average number of working hours per week for full-time workers in April 1997 is 45.79 (with a standard deviation of 9.2 h). It is noteworthy that Card (1999) found that education had a positive impact on both hours of work per week and the number of weeks worked per year. The current analysis examines how individuals and institutional characteristics are related to hours of work per week and implicitly assumes the same number of weeks worked per year across individuals, thus it provides a conservative estimate for the impact of various factors on hours of work per year.

This analysis also employs a variety of independent variables that have been shown in previous studies to affect graduates’ earnings. These variables include various demographic characteristics, family backgrounds, college characteristics, academic backgrounds, and labor market variables. Demographic variables consist of gender and race/ethnicity, consisting of several dummy variables such as “Female,” “White,” “Indian American,” “Asian,” “Black,” and “Hispanic.” Family background variables provide the socioeconomic status of these college graduates through two main variables: family income and whether he or she is a first-generation college graduate. The latter captures the educational attainment of the college graduates’ parents. When these two variables are inserted in the earnings equation, other characteristics of socioeconomic status such as parents’ professions do not seem to influence college graduates’ earnings significantly, so this study does not consider these variables.

The variables measuring institutional selectivity are constructed from the Integrated Postsecondary Education Data System 1992–93 (IPEDS) and Barron’s Profiles of American Colleges. Specifically, I use dummy variables to indicate six college types: high-quality private, high-quality public, middle-quality private, middle-quality public, low-quality private, and low-quality public institutions. Detailed procedures of constructing these variables have been discussed in several previous studies including Brewer and Ehrenberg (1996), Brewer et al. (1999), and Thomas and Zhang (2005).

Academic variables include college majors (divided into business, education, engineering, health sciences, public affairs, biological science, social science, math sciences, history, the humanities, psychology, and other majors) and student academic performance (measured by undergraduate GPA). One special set of academic variables is post-baccalaureate degree attainment. Graduate education has increasingly become an integral part of human capital accumulation. In the literature of the relationship between college education and labor market outcomes, graduate education is especially important because it may serve as an intermediate link. For example, Zhang (2005) examines the effect of college quality and undergraduate majors on a variety of graduate education outcomes including graduate school enrollment, graduate degree attainment, and the quality of graduate programs, and finds that college quality, among other academic factors, has a significant effect on post-baccalaureate degree attainment. Considering that an advanced degree is usually perceived as a prerequisite to many desirable and prestigious professions with great economic rewards and high social status (e.g., physician, professor, lawyer, and scientist), post-baccalaureate degree attainment should have a significant impact on graduates’ earnings. However, because working experience and job-specific training are also important during early careers of college graduates, a relatively long period of time is probably needed to observe any significant impact of graduate education. The three waves of B&B data provide a unique opportunity to examine these issues.

Sample

The starting sample size is 10,080, 10,093, and 8,969 for the 1994, 1997, and 2003 B&B follow-up surveys. I took the full sample from each survey and limited it to graduates (a) who received their first bachelor’s degrees between July 1992 and June 1993, (b) who were working full-time when the follow-up survey was conducted, and (c) who had reported valid earnings data. More specifically, I excluded individuals who reported hourly earnings that were less than the federal minimum wage rate ($4.25 for 1994 and $5.15 for 1997 and 2003). I also excluded individuals with annual incomes of more than $500,000. These criteria limited the sample size to 5,561, 6,426, and 5,765 for the 1994, 1997, and 2003 samples, respectively.4 In addition, when institutional-level data are merged with the B&B data, I further excluded individuals with missing institutional-level data. This step limited the final sample size to 5,154, 5,965, and 5,343 for the 1994, 1997, and 2003 samples, respectively. It is noteworthy here that because list-wise deletion was used to cope with missing data in subsequent analysis, the actual number of observations in regression equations also depended on the number of missing data in each variable included in the analysis.

Estimation

There are two steps in the empirical analyses. In the first step, I run a series of hierarchical linear models (HLM) to examine the impact of various individuals and institutional factors on graduates’ earnings and hours of work. Due to the multilevel nature (i.e., institutional and individual) of factors shown to have effects on the outcome of interest (i.e., earnings and hours of work) in the current analysis, econometric techniques which characterize this nature such as hierarchical linear modeling (HLM) are often recommended (Bryk and Raudenbush 1992; Heck and Thomas 2000). In this particular analysis, the effect of individual characteristics is restricted to be the same across institutions. That is, I estimate a series of intercepts-as-outcome models:
$$ \ln ({\mathop Y\nolimits_{ij} })\, = \,{\mathop \gamma \nolimits_{00} }\, + \,{\mathop \gamma \nolimits_{01} }{\mathop Z\nolimits_j }\, + \,{\mathop \gamma \nolimits_{10} }{\mathop X\nolimits_{ij} }\, + \,{\mathop \mu \nolimits_j }\, + \,{\mathop \varepsilon \nolimits_{ij} } $$
(1)
where Yij represents individuals’ earnings and hours of work, Zj is the quality of institution j he or she actually attended, and Xij is the matrix of individual-level variables including demographic characteristics, family background, academic background, and job market conditions.
In the second step, the impact of various factors on actual realized hourly income is derived from the first step analyses. By construction, the impact on hourly income is the difference between the impact on earnings and on hours of work. It holds that hourly income (w) is the division of earnings (Y) and hours of work (H)5
$$ w\, = \,Y/H $$
(2)
After taking logarithms on both sides, one obtains
$$ \ln w\, = \,\ln Y\, - \,\ln H $$
(3)
It follows that the effect of any variable (x) on hourly income is the difference between its effect on the (logged) earnings and its effect on (logged) hours of work.
$$ \frac{{\partial \ln w}} {{\partial x}}\, = \,\frac{{\partial \ln Y}} {{\partial x}}\, - \,\frac{{\partial \ln H}} {{\partial x}} $$
(4)
where x represents each of the independent variables in Eq. 1.

Results

The estimation of HLM usually consists of two steps. The first step is to estimate a model with no explanatoty variables in order to partition the total variance in outcomes within and between institutions in the sample. The model is usually referred to as the “null model” or one-way ANOVA model. The information of the variance components is used to determine whether an HLM analysis is necessary. The result of this simple one-way ANOVA analysis is presented in Table 1. For example, about 10–11% of total variance in (log) earnings is between institutions, the within-institution variance makes up the remaining. The proportion of between-institution variance is smaller at approximately 2–3% for (log) hours of work. Although the variance between institutions is relative small, it is conceptually important to model how institutional factors might affect individual outcomes. In addition, HLM takes the nested structure of the data into account when computing the standard errors in regression analysis.
Table 1

Variance components of dependent variables, by year

 

1994

1997

2003

Log earnings

Variance within institutions

0.1437

0.1500

0.2134

Variance between institutions

0.0191

0.0203

0.0239

Proportion between institutions

0.1173

0.1194

0.1009

Log hours of work

Variance within institutions

0.0229

0.0277

0.0301

Variance between institutions

0.0007

0.0005

0.0006

Proportion between institutions

0.0290

0.0177

0.0191

The results of HLM regression analyses are presented in Tables 2 and 3. Table 2 reports regression results for college graduates’ earnings in 1994, 1997, and 2003, and Table 3 presents results for hours of work. For each regression equation, independent variables are arranged in several blocks, namely, institutional characteristics, demographic characteristics, family background, academic background, and, for the last two survey samples, post-baccalaureate degree attainment. Not reported but included in each regression equation are the individual’s age at the time of survey interview and its squared terms, and the individual’s tenure at the current job at the time of survey interview and its squared terms. From Eq. 4, the impact of these variables on the hourly income can be computed as the difference between the regression coefficients for earnings and hours of work equations or more straightforwardly estimated by using hourly wage rate as the dependent variable. These results are reported in Table 4.
Table 2

HLM estimates for college graduates’ earnings (t-statistics in parentheses)

Variable

1994

1997

2003

Coeff.

t-value

Coeff.

t-value

Coeff.

t-value

Constant

8.9656

(81.06)

9.2949

(70.15)

10.8513

(44.49)

Institutional characteristics

Low-quality, private institution

−0.0333

(−1.14)

−0.0046

(−0.15)

0.0132

(0.35)

Middle-quality, public institution

0.0118

(0.70)

0.0370

(2.06)

0.0646

(3.03)

Middle-quality, private institution

0.0063

(0.33)

0.0287

(1.47)

0.0505

(2.12)

High-quality, public institution

0.0837

(2.48)

0.1315

(3.59)

0.1663

(3.99)

High-quality, private institution

0.0644

(2.43)

0.1532

(5.47)

0.1901

(5.83)

Historically Black colleges & univ.

−0.0855

(−1.93)

−0.0621

(−1.46)

−0.0202

(−0.38)

Demographic characteristics

Female

−0.1100

(−10.28)

−0.1460

(−13.87)

−0.2217

(−16.79)

Indian American

0.0236

(0.41)

0.0139

(0.24)

0.1052

(1.37)

Asian

0.0261

(0.90)

0.1097

(3.84)

0.0081

(0.23)

Black

0.0038

(0.15)

−0.0065

(−0.27)

−0.0318

(−1.03)

Hispanic

−0.0202

(−0.80)

−0.0148

(−0.61)

−0.0348

(−1.10)

Family background

Family income (in $10,000)

0.0038

(3.60)

0.0055

(4.83)

0.0077

(5.90)

First-generation college graduate

−0.0027

(−0.26)

−0.0056

(−0.56)

−0.0313

(−2.44)

Academic background

Undergraduate GPA

0.0686

(6.50)

0.0623

(5.94)

0.0562

(4.22)

Business major

0.1928

(10.56)

0.2495

(13.89)

0.2743

(11.77)

Engineering major

0.3790

(15.93)

0.3824

(16.57)

0.3509

(12.24)

Health major

0.3970

(17.79)

0.3332

(15.10)

0.3363

(11.66)

Public affair major

0.0627

(2.30)

0.0939

(3.54)

0.0820

(2.37)

Biological science major

0.0569

(2.07)

0.1073

(3.82)

0.2075

(6.57)

Math science major

0.2311

(9.59)

0.2593

(11.27)

0.2476

(8.78)

Social science major

0.0775

(3.80)

0.1474

(7.44)

0.1862

(7.44)

History major

0.0234

(0.61)

0.0629

(1.73)

0.0850

(1.94)

Humanity major

0.0163

(0.77)

0.0514

(2.51)

0.0536

(2.03)

Psychology major

−0.0186

(−0.66)

0.0511

(1.83)

0.0709

(2.00)

Other major

0.0843

(4.60)

0.1381

(7.72)

0.1511

(6.61)

Post-B.A. degrees

Master’s and first professional

  

0.0544

(3.46)

0.1075

(7.48)

PhD

  

−0.0152

(−0.46)

0.0693

(1.82)

N

4936

 

5717

 

5060

 

Variance within institutions

0.1134

 

0.1262

 

0.1790

 

Variance between institutions

0.0061

 

0.0084

 

0.0093

 

Note: Also included in the model are the individual’s age and its squared term, tenure at the current job and its squared term

Table 3

HLM estimates for college graduates’ hours of work (t-statistics in parentheses)

Variable

1994

1997

2003

Coeff.

t-value

Coeff.

t-value

Coeff.

t-value

Constant

3.7943

(80.22)

3.6739

(63.52)

3.8156

(40.74)

Institutional characteristics

Low-quality, private institution

−0.0043

(−0.37)

0.0014

(0.13)

−0.0096

(−0.75)

Middle-quality, public institution

−0.0018

(−0.28)

−0.0040

(−0.68)

0.0088

(1.36)

Middle-quality, private institution

−0.0073

(−0.98)

0.0015

(0.22)

0.0078

(1.01)

High-quality, public institution

0.0186

(1.49)

0.0132

(1.22)

0.0048

(0.41)

High-quality, private institution

−0.0021

(−0.20)

0.0347

(3.56)

0.0432

(4.17)

Historically Black colleges & univ.

0.0116

(0.63)

0.0115

(0.69)

−0.0105

(−0.55)

Demographic characteristics

Female

−0.0556

(−12.06)

−0.0683

(−14.66)

−0.0642

(−12.57)

Indian American

−0.0143

(−0.58)

0.0217

(0.85)

0.0090

(0.30)

Asian

−0.0299

(−2.42)

−0.0257

(−2.08)

−0.0033

(−0.25)

Black

−0.0469

(−4.26)

−0.0332

(−3.13)

−0.0182

(−1.51)

Hispanic

−0.0321

(−2.98)

−0.0256

(−2.44)

−0.0249

(−2.08)

Family background

Family income (in $10,000)

0.0003

(0.67)

0.0022

(4.28)

0.0015

(2.93)

First-generation college graduate

−0.0070

(−1.57)

−0.0080

(−1.80)

−0.0122

(−2.46)

Academic background

Undergraduate GPA

−0.0100

(−2.21)

−0.0055

(−1.19)

0.0004

(0.07)

Business major

0.0195

(2.48)

0.0226

(2.85)

0.0032

(0.35)

Engineering major

0.0019

(0.19)

0.0076

(0.76)

−0.0191

(−1.76)

Health major

−0.0397

(−4.18)

−0.0389

(−4.11)

−0.0262

(−2.38)

Public affair major

−0.0214

(−1.83)

−0.0275

(−2.34)

−0.0361

(−2.69)

Biological science major

−0.0124

(−1.05)

−0.0023

(−0.18)

0.0281

(2.30)

Math science major

−0.0093

(−0.90)

−0.0060

(−0.59)

−0.0133

(−1.22)

Social science major

−0.0057

(−0.65)

0.0053

(0.61)

−0.0100

(−1.03)

History major

0.0103

(0.62)

0.0203

(1.26)

0.0061

(0.36)

Humanity major

−0.0134

(−1.47)

−0.0017

(−0.18)

−0.0021

(−0.21)

Psychology major

−0.0321

(−2.62)

−0.0386

(−3.11)

−0.0202

(−1.47)

Other major

0.0108

(1.37)

0.0193

(2.45)

0.0100

(1.14)

Post-B.A. degrees

Master’s and first professional

  

−0.0044

(−0.62)

0.0293

(5.23)

PhD

  

0.0160

(1.08)

0.0583

(3.93)

N

4936

 

5717

 

5060

 

Variance within institutions

0.0214

 

0.0259

 

0.0280

 

Variance between institutions

0.0005

 

0.0000

 

0.0000

 
Table 4

HLM estimates for college graduates’ hourly income (t-statistics in parentheses)

Variable

1994

1997

2003

Coeff.

t-value

Coeff.

t-value

Coeff.

t-value

Constant

1.2292

(11.60)

1.6650

(13.15)

3.0948

(12.74)

Institutional characteristics

Low-quality, private institution

−0.0298

(−1.07)

−0.0074

(−0.25)

0.0217

(0.58)

Middle-quality, public institution

0.0136

(0.83)

0.0415

(2.38)

0.0554

(2.65)

Middle-quality, private institution

0.0123

(0.68)

0.0257

(1.36)

0.0404

(1.73)

High-quality, public institution

0.0632

(1.94)

0.1160

(3.25)

0.1558

(3.84)

High-quality, private institution

0.0659

(2.58)

0.1198

(4.42)

0.1470

(4.60)

Historically Black colleges & univ.

−0.0986

(−2.31)

−0.0770

(−1.88)

−0.0119

(−0.22)

Demographic characteristics

Female

−0.0544

(−5.30)

−0.0771

(−7.67)

−0.1576

(−11.98)

Indian American

0.0387

(0.71)

−0.0057

(−0.10)

0.0953

(1.25)

Asian

0.0533

(1.92)

0.1312

(4.81)

0.0071

(0.21)

Black

0.0512

(2.09)

0.0283

(1.23)

−0.0119

(−0.39)

Hispanic

0.0106

(0.43)

0.0121

(0.52)

−0.0090

(−0.29)

Family background

Family income (in $10,000)

0.0036

(3.52)

0.0035

(3.23)

0.0063

(4.84)

First-generation college graduate

0.0040

(0.40)

0.0017

(0.18)

−0.0195

(−1.53)

Academic background

Undergraduate GPA

0.0778

(7.70)

0.0678

(6.76)

0.0551

(4.15)

Business major

0.1728

(9.88)

0.2280

(13.29)

0.2726

(11.74)

Engineering major

0.3771

(16.53)

0.3768

(17.08)

0.3726

(13.05)

Health major

0.4362

(20.38)

0.3677

(17.43)

0.3632

(12.64)

Public affair major

0.0835

(3.20)

0.1219

(4.82)

0.1181

(3.42)

Biological science major

0.0690

(2.62)

0.1097

(4.09)

0.1787

(5.68)

Math science major

0.2399

(10.39)

0.2661

(12.12)

0.2624

(9.34)

Social science major

0.0824

(4.22)

0.1423

(7.51)

0.1976

(7.92)

History major

0.0151

(0.41)

0.0471

(1.36)

0.0839

(1.93)

Humanity major

0.0294

(1.45)

0.0542

(2.77)

0.0557

(2.12)

Psychology major

0.0127

(0.47)

0.0919

(3.43)

0.0905

(2.55)

Other major

0.0729

(4.15)

0.1192

(6.98)

0.1427

(6.27)

Post-B.A. degrees

Master’s and first professional

  

0.0597

(3.97)

0.0792

(5.53)

PhD

  

−0.0325

(−1.02)

0.0115

(0.30)

N

4936

 

5717

 

5060

 

Variance within institutions

0.1040

 

0.1148

 

0.1784

 

Variance between institutions

0.0058

 

0.0083

 

0.0084

 

First, turning to the earnings equations in Table 2, the estimated effect of various variables on earnings for the 1994 and 1997 sample confirms the results from Thomas and Zhang’s (2005) earlier study that uses a consistent sample in 1994 and 1997 to examine the changing impact of college education on graduates’ earnings. Net of all other variables in the model, the effects of college quality are small although statistically significant in 1994 (column 1). Relative to graduates from low-quality public institutions (the comparison group in each model), graduates from middle-quality and high-quality colleges enjoy a roughly 6–8% earnings advantage. While the economic returns to college quality are relatively small shortly after graduation, considerable differences do emerge several years later in 1997. The results in column 2 show that graduates from high-quality public and private colleges enjoy a 14% and 16% (log coefficients of 0.132 and 0.152) earnings advantage relative to graduates from low-quality public colleges.6 The significant earnings difference among graduates from different institutions seems to persist 10 years after college graduation. The earnings equation for the 2003 sample shows that graduates from high-quality public and private institutions enjoy an earnings premium of 18% and 20% (log coefficients of 0.166 and 0.190) relative to graduates from low-quality public institutions.

Regression estimates have also revealed earnings differences associated with demographic characteristics. For all three samples, female graduates are consistently found to earn less than their male counterparts, and the gender gap increases from 10% in 1994 to 13% in 1997 and to 20% in 2003 (−0.110, −0.146, and −0.222 log points, respectively). The earnings differences among between race/ethnicity do not appear to be significant, holding other factors in the regression constant.

Family background plays a role in earnings at all three time points. Family income is positively related to earnings for all three samples, although the impact is relatively small. Being a first-generation college graduate is negatively associated with earnings only in 2003, but not at the earlier two time points. The significant but relatively small impact of family background variables on graduate earnings is consistent with previous studies that examined the direct and indirect impact of family background on students’ outcomes. Generally speaking, the family backgrounds tend to have an indirect impact on earnings through educational attainment rather than a direct impact.

Among all variables included in the model, undergraduate majors appear to have the greatest leverage on graduate earnings. For all three samples, graduates from fields in business, engineering, health, and math/science enjoy significant earnings advantages over their peers in other major fields of study; however, there seems to be some differences in the earnings trajectories across majors. The earnings advantages for some majors seem quite stable over time. For example, graduating with a degree in engineering or health field provides about 46–48% (0.379 and 0.397 log points) earnings advantages relative to education majors about 1 year after graduation. At the time of the third follow-up survey, these earnings advantages were 42% and 40% (0.350 and 0.336 log points). In contrast, other majors such as business and biological sciences have increased their advantages over education majors substantially over time.

The analysis of the impact of post-baccalaureate degree attainment on earnings yields some interesting results. Post-baccalaureate degree attainment is not included in 1994 because few students obtained advanced degree within 1 year after college graduation. Results in 1997 indicate that college graduates who obtained master’s degrees within 4 years after college graduation enjoyed about a 6% earnings premium over those who did not. Because very few students have completed their doctoral degrees within 4 years after college graduation, the analysis of the impact of PhD degree completion on earnings yields insignificant results. When earnings data are collected 10 years after college graduation, however, the earnings advantage enjoyed by master’s degree holders has increased to 11% (0.108 log points) while for those who completed their PhDs, an earnings advantage has not been evident yet.

Table 3 presents the impact of various factors on hours of work. In general, these models have low predictive power (i.e., low proportion of variances explained); however, important exceptions emerge across these equations. The most noticeable exception to the impact of college characteristics on hours of work is longer hours of work for graduates from high-quality private institutions. For example, the second column shows that graduates from high-quality private institutions, on average, spend 3.5% more time working in 1997 than graduates from other types of institutions. A similar gap of 4.4% is observed in the last column when hours of work are evaluated in 2003. Because of longer hours of work by graduates from high-quality private institutions, the advantage in terms of hourly income enjoyed by these graduates relative to their peers from other types of institutions is significantly less than the advantage in terms of earnings. For example, the earnings advantage of 16% (Table 2, column 2, 0.153 log points) enjoyed by graduates from high-quality private institutions relative to those from low-quality public institutions in 1997 is reduced to about 12% (0.119 log points) when the hourly income is evaluated (Table 4, column 2). And in 2003, the 21% (Table 2, column 3, 0.190 log points) earnings advantage is reduced to a 15% (Table 4, column 3, 0.147 log points) advantage in hourly income.

Fewer hours of work by female graduates account for a significant proportion of the gender gap in earnings. For example, the gender gap in earnings is estimated at 10%, 13%, and 20% in 1994, 1997, and 2003, respectively (Table 2). Table 3 indicate that about 6–7% of the gap is attributable to fewer hours of work by female graduates, leaving the gender gap in hourly income at 5%, 7%, and 14% in 1994, 1997, and 2003, respectively (Table 4). I do not attempt to explain why full-time female workers spend less time working; however, hours of work seems to be an important factor for the gender gap in earnings.

Family backgrounds appear to influence graduates’ hours of work. Graduates from high-income family tend to work longer hours, which reduces the small effect of family income on graduates’ earnings further when the hourly income is evaluated. First-generation college graduates tend to work 1–2% fewer hours, a small but statistically significant difference. Table 3 does not detect any significant relationships between individuals’ undergraduate GPAs and hours of work. As a result, the large and significant relationship between GPA and earnings is reflected in the relationship between GPA and hourly income (Table 4). The variation in hours of work across undergraduate majors does not seem to be as large as expected. The majority of the estimated differences between education majors and other majors are within ±3% range and statistically insignificant. Probably the most consistent result is the relatively longer hours of work by business majors and relatively shorter hours of work by health and public affair majors.

One empirical extension of the analyses warrants being briefly reported here. First, male and female graduates might have different choices of jobs and hours of work because the latter assume greater family responsibilities. The analyses show that, on average, female graduates work 6–7% fewer hours than their male counterparts. To examine whether male and female graduates have different determination structure for earnings and hours of work when they are employed full-time, I replicate all analyses in Tables 23 for male and female workers separately. Results suggest that in general the conclusion from the pooled sample holds for male and female groups.

Discussion and Conclusion

This study extends the analysis of the economic return of college education up to 10 years after college education and further examines the impact of college education on graduates’ hours of work. The results suggest that variation in hours of work accounts for a small portion of earnings differences among college graduates; however, some important findings emerge from this analysis. Graduates from high-quality private institutions tend to work longer hours than their peers from other types of institutions. Female graduates spend fewer hours working than their male counterparts. As far as family background is concerned, graduates from high-income families tend to work longer hours and first-generation college graduates tend to work fewer hours. Finally, business majors seem to work longer hours while health and public affair majors less hours.

These results provide another layer of important information to students in making decisions about college education. This study is based on the premise that individuals’ decision should be guided by the utility maximization principle, rather than a single economic outcome such as earnings. The utility framework that underlies this study is different from the human capital framework, which ultimately compares the costs and benefits associated with different educational choices. Under the utility framework, individuals’ preferences are introduced into the decision process. For example, it is entirely possible that attending a private elite institution is a good choice under the human capital framework but not under the utility framework because the latter takes the positive utility of leisure into consideration. Because individuals have difference preferences about earnings and leisure, the results of current analyses may lead to different choices for different individuals. For example, although relatively high earnings enjoyed by graduates from private elite institutions might yield a high satisfaction level overall, the fact that they work longer hours might result in lower job satisfaction. The longer hours of work by those from high-quality private institutions seems to provide a possible explanation for their relatively low job satisfaction (Zhang and Thomas 2003).

Caution must be taken in interpreting the results of this study. For example, one of the main findings of the current study is that students who graduate from high-quality institutions have higher hourly income and at the same time work longer hours. The standard explanation for this result is that individuals who enjoy higher hourly income have more incentives to work longer hours. Yet, they could choose to work less. If that is the case, there seems to be some option values associated with having a degree from high-quality institutions because it is conceivable that those students from high-quality institutions have more flexibility in determining the balance between their work and leisure time than would their counterparts from other schools. However, without information on options available to graduates from different types of institutions and their choices, it is difficult to test to what extent college quality affects graduates’ choices in the labor market. Beside this standard interpretation, there could be other possibilities too. For example, students who choose to attend private elite institutions might be more ambitious, which could also lead to longer hours of work once they are in the labor market. That is, the relationship between attending private elite institutions and longer hours of work could be driven by certain underlying individual characteristics. Although the current analysis included a variety of individual characteristics, it is difficult to rule out this possibility without knowing why students make different choices in the college and in the labor market.

Several important questions are left unanswered in this study. The earnings of college graduates increase rapidly during their first 10 years after college graduation. The average earnings for the sample of full-time workers are $24,706, $35,028, and $60,443 in 1994, 1997, and 2003, respectively, averaging more than 10% increase every year. (The median values are $22,490, $31,000, and $52,000). On the other hand, the hours of work also increased from 44.01 h per week in 1994 to 45.79 h in 1997 and to 46.82 h in 2003. It would be important to know what causes this fast salary increase, and how college education is related to job opportunities such as promotion, further education and training, and career mobility. While the current study examines the variation in earnings and hours of work for full-time workers, other job market outcomes could potentially be influenced by college education. For example, the probability of being unemployed or employed part-time might be related to various education and non-educational factors. Future research needs to go beyond earnings differentials among full-time workers to provide a more complete view about the impact of college education on labor market outcomes.

Footnotes
1

This static labor supply model is discussed extensively in the literature of labor economics. See, for example, Blundell and MaCurdy (1999), for a review of different models of labor supply and the elasticity of individual and aggregate labor supply. See also MaCurdy et al. (1990) for a detailed discussion on the empirical evidence and methods used to estimate the impact of wage rate on hours of work.

 
2

To examine the dynamics between the wage rate and hours of work, I estimate two models. In the first model, I restrict the sample to those full-time workers who reported their earnings in terms of hourly wage rates. For example, 783 out of the 6,426 individuals in the 1997 sample did so. Adding the hourly wage rate as an additional independent variable to the hours of work equation (which is discussed in next section) verifies that as the hourly wage rate increases, individuals tend to work more hours. In the second model, I use the whole sample of full-time workers and added the hourly income in the hours of work equation, its coefficient turned out to be negative, which is not a surprise given the relationship between earnings, hours of work, and hourly income and the possibility of division bias (Borjas 1980). These findings confirmed the analytical approach of the current study, i.e., the hourly income is not assumed to be predetermined; instead, it is viewed as the realized income per hour, which is determined by the salary and hours of work.

 
3

Restricted-use data of B&B study are available from NCES. More information about NCES data licensing and other issues is available at http://www.nces.ed.gov

 
4

The selection criteria here are more inclusive than those used in Thomas and Zhang (2005). For example, their study excluded students with subsequent degrees after their B.A. completion, while this analysis includes dummy variables to indicate post-B.A. degree completion.

 
5

Computing wage rate this way might create problems when there is measure error in reported hours of work in empirical analysis. See for example, Borjas (1980) for a detailed discussion on this issue. However, because the current analysis examines the relationship between hourly income and other individual and institutional factors, the possible measurement error in hours of work is less a problem than in studies that examine the relationship between wage rate and hours of work.

 
6

These coefficients are slightly different from Thomas and Zhang (2005) because different sample criteria and model specifications are used; however, the qualitative results are same.

 

Copyright information

© Springer Science+Business Media, LLC 2007