293 A. Light Economics of Education Review 18 1999 291–309
time-varying nature of the wage benefits associated with high school employment.
Third, I hold constant postsecondary schooling attain- ment and post-school work experience, which allows me
to interpret my estimates using a standard, human capital framework. Ruhm observes career outcomes 6–9 years
after high school, at which time respondents differ dra- matically with respect to their postsecondary schooling
and work histories. He omits measures of schooling attainment and post-high school work experience from
his set of covariates because they are endogenous, but as a result he cannot separate the productivity-enhancing
effects of high school employment from the effects of subsequent schooling attainment and on-the-job train-
ing.
3
I eliminate heterogeneity in postsecondary school- ing by limiting my sample to terminal high school gradu-
ates, and I control for heterogeneity in post-school, on- the-job training with measures of actual and potential
work experience. As noted above, I then contend with the endogeneity of work experience using an instrumen-
tal variables approach.
A final difference between this study and its prede- cessors is that I explicitly consider the potential trade-
off between high school employment and high school achievement. To control for high school achievement in
my wage models, I use data on subject-specific credit hours earned by each respondent—information that is
available because high school transcripts were collected and coded for a large number of NLSY respondents.
Detailed transcript data are indispensable if we wish to assess the effect of high school employment on wages
net of its effect on skills being learned contempor- aneously inside the classroom. In particular, these data
enable us to investigate the presumption that employed high school students shun academic subjects in favor of
such courses as typewriting, auto mechanics, and choir— curricula choices that may leave them with ample free
time and even high grade point averages, but with poorer post-school wage earning capabilities than their nonem-
ployed counterparts. This premise receives support in a recent study by Eckstein and Wolpin 1998, who use
NLSY data to find a small, negative relationship between
3
Ruhm’s approach is akin to identifying the returns to schooling with a wage model that omits post-school experience
from the covariates. Because schooling and work experience are strongly, positively correlated, such a model would imply
much larger returns to schooling than are found in conventional models. To circumvent this difficulty, Ruhm presents one speci-
fication in which the sample is confined to individuals who average at least 1000 h or 26 weeks of work during the period
in which earnings are observed 6–9 years after high school. This is only a partial solution, for it still allows considerable
variation in work effort within that window, as well as in the preceding years; in addition schooling attainment remains
uncontrolled for.
high school work effort and the accumulation of credit hours. Neither Ruhm nor Hotz et al. include measures
of high school achievement among their regressors, although a number of earlier studies use grade point
averages or class rank as outcome measures in assessing the value of high school employment.
In the next section I explain how I select the sample of male high school graduates used throughout the study,
and I provide a descriptive analysis of these respondents’ characteristics. In particular, I describe the extent of their
high school employment experiences and summarize the relationships between high school employment and
numerous other characteristics, including high school achievement. In Section 3 I describe the wage models
to be estimated, define the covariates, and discuss the estimation technique. Section 4 presents the estimates,
and in Section 5 I offer concluding remarks.
2. Data
2.1. Sample selection The data are from the National Longitudinal Survey
of Youth NLSY, which began in 1979 with a sample of 12,686 males and females born in 1957–64. Respondents
were interviewed annually from 1979 to 1994, at which time the survey became biennial. I use data from inter-
view years 1979–91, and I restrict the analysis to a sub- sample of 685 males. Gender differences in high school
employment are documented in Michael and Tuma 1984, Light 1995a, and Ruhm 1997, so in this paper
I opt to focus exclusively on male workers. Further reductions in sample size are caused primarily by my
requirements that a high school transcript data be avail- able for each respondent, b their labor market experi-
ences be “observed” from the start of grade 11 onward, and c they receive no postsecondary schooling. In the
remainder of this subsection I elaborate on the selection criteria and also describe the NLSY transcript and
employment data.
4
In 1980, 1981, and 1983 an attempt was made to col- lect high school transcripts for eligible NLSY respon-
dents. Respondents were eligible if they consented to release their high school records, did not attend high
schools outside the United States, and were not in the military subsample.
5
In addition, respondents had to
4
Additional information on the data can be found in US Department of Labor 1998 and Light 1995b.
5
The original NLSY sample of 12,686 respondents consisted of a nationally representative subsample n 5 6111, an over-
sample of Hispanics, blacks and economically disadvantaged whites n 5 5295, and a military subsample of individuals who
were enlisted in the military on or before 30 September 1978 n 5 1280.
294 A. Light Economics of Education Review 18 1999 291–309
graduate from or otherwise exit high school before their transcripts were collected. Although more than 10,000
NLSY respondents were eligible for the transcript collec- tion in one of the three years, completed transcripts were
acquired and coded for only 9010 respondents. I make an initial reduction in sample size by eliminating the
6283 female respondents, and also deleting 824 male members of the military subsample and an additional
1150 men for whom transcript data are unavailable. These deletions reduce the sample size to 4429.
I delete an additional 1082 men because they did not graduate from high school. Although the relationship
between high school employment and the likelihood of graduation is a worthy subject for analysis see D’Am-
ico, 1984; Marsh, 1991; Eckstein and Wolpin, 1998, I choose to exclude high school dropouts so I can look at
all respondents’ in-school employment experiences for two academic years prior to their school exit. If dropouts
were included in the analysis, I would have to contend with the fact that they are legally barred from holding
a job for most or perhaps all of their last two years of school.
The NLSY asks respondents about their work experi- ences with virtually every employer encountered from
January 1978 onward or, for respondents who were not yet age 16 at that date, from age 16 onward. The reported
information is used to create week-by-week variables on labor force status and hours worked on all jobs in pro-
gress during the given week.
6
To ensure that each respondent’s employment experiences are recorded from
the start of grade 11 onward, I delete respondents from the sample if this date precedes January 1978; with few
exceptions, all remaining respondents graduate from high school in the spring of 1980 or later. This leads to
an additional 1561 respondents being dropped, and reduces the sample size to 1786.
I eliminate an additional 192 individuals because their transcripts are incomplete for grade 11 or 12. I require
each respondent’s transcript to show at least four courses or four Carnegie credits for grades 11 and 12, where
one Carnegie credit represents a year-long course.
7
I also eliminate 27 respondents who fail to report any employ-
ment experiences during the 9 years following high school graduation. By observing career outcomes over a
9-year window I am able to identify changes over time in the wage effect of high school work experience. I choose
9 years as my cut-off because it corresponds to Ruhm’s 1997 6–9 year outcome window, and because survey
6
The arrays of weekly data are released as part of the NLSY work history file.
7
As part of the transcript collection effort, courses were assigned three-digit codes identifying their content and credit
hours were translated into Carnegie units. This was intended to produce uniformity across schools.
attrition reduces the number of respondents seen more than 9 years after their high school graduation date. A
small number of respondents drop out of the NLSY prior to reaching the 9-year mark, in which case I simply
observe them until their last interview date.
The deletion rules described above produce a sample of 1567 male respondents who graduate from high
school and whose employment histories are recorded from the start of grade 11 onward. Many of these indi-
viduals proceed to attend college and hold college jobs. The intervention of college attendance and employment
complicates my efforts to identify the relationship between high school employment and subsequent wages,
so I confine the sample to 685 terminal high school graduates—that is, respondents who report no postsec-
ondary enrollment during the observation period. NLSY respondents are asked their enrollment status on a
month-by-month basis from January 1980 onward and in a more general manner prior to that date, so I can
accurately identify even very short postsecondary enrollment spells for the respondents in my sample. I
have replicated the entire analysis for the larger sample of 1567 respondents, and I summarize the key differ-
ences between the two groups in subsequent sections.
2.2. Sample characteristics In this subsection, I describe the amount of employ-
ment experience acquired by the 685 sample members while enrolled in grades 11 and 12. I then segment the
sample by high school work intensity and examine dif- ferences across groups in a large number of character-
istics, including those related to high school achieve- ment. This descriptive analysis reveals a number of
interesting contrasts between individuals who choose dif- ferent levels of high school employment effort, and it
sets the stage for the regression analysis presented in subsequent sections.
8
Table 1 describes the distribution of average hours worked per week during the 2 years preceding high
school graduation. The 2-year period is broken down into four contiguous segments: the summer before grade 11,
the grade 11 academic year, the summer before grade 12, and the grade 12 academic year. To measure average
weekly hours of work, I count the total number of hours worked during each interim, and divide by the number
of weeks elapsed. The average and modal durations of each academic year and summer are 36 weeks and
14 weeks, respectively. Each respondent’s employment history must be recorded from the onset of grade 11 in
8
Most of the characteristics summarized in this section are used as covariates in the wage model described in Section 3.
Details on how the variables are defined and constructed are deferred to Section 3.
295 A. Light Economics of Education Review 18 1999 291–309
Table 1 Distribution of average hours worked per week in high school
Percent of individuals Average hours worked per week
Summer before Grade 11
Summer before Grade 12
Grades 11 and 12 grade 11
grade 12 27.9
41.6 26.7
29.3 22.8
1–10 37.4
24.8 15.9
20.3 34.7
11–20 14.7
17.4 22.3
20.0 25.1
21 1 20.0
16.2 35.0
30.4 17.4
Mean 12.3
8.5 15.7
13.4 10.3
S.D. 13.1
10.9 14.1
13.2 10.0
Mean among workers 19.5
14.5 21.3
19.0 13.0
S.D. 11.5
10.7 12.3
11.9 9.6
Number of individuals 515
685 685
685 685
order for him to remain in the sample, but employment during the summer preceding grade 11 is unknown for
25 of the sample. Table 1 reveals that many high school juniors and
seniors work a substantial number of hours during the academic year. The column titled “Grade 11” reveals that
41.6 of respondents do not work during their junior year of high school, while 24.8 average 1–10 h per
week, 17.4 average 11–20 h per week, and the remain- ing 16.2 work more than 20 h per week during the aca-
demic year. As one would expect, employment is even more prevalent during the senior year of high school.
The column titled “Grade 12” indicates that only 29.3 of sample members do not work during their senior year
of high school, while 30.4 average over 20 h per week; the remaining respondents are evenly divided between
working 1–10 and 11–20 h per week. The average num- ber of hours worked per week increases from 8.5 in
grade 11 to 13.4 in grade 12; among those students who work, the average weekly effort increases from 14.5 h in
grade 11 to 19.0 h in grade 12.
9
Table 1 reveals the temporal patterns one expects to see in the employment of young people. In addition to a
pronounced rightward shift in the distribution of average hours between the grade 11 and grade 12 academic
years, there is also a rightward shift between the summer before grade 11 and the summer before grade 12. Young
men are more likely to be employed as they age and,
9
The patterns seen in Table 1 change very little when I use a larger sample of 1567 men that includes college-goers. The
only difference is that college-goers are slightly less likely than terminal high school graduates to work very intensively. In
grade 12, for example, only 27.8 of the larger sample aver- ages 21 or more hours per week of employment, and the aver-
age work effort among workers falls from 19 to 17.7 h per week.
conditional on working, tend to increase their hours. However, there is not a monotonic increase in work
effort over time, for high school students frequently work intensively during the summer and then cut back on their
hours or quit their jobs altogether during the sub- sequent academic year. Nonetheless, it is relatively rare
for young men to remain nonemployed throughout their last two academic years of high school. The right-most
column of Table 1 shows that only 23 of the sample gains no work experience while enrolled in grades 11
and 12, while the typical student averages 10 h of work per week during this period of time.
In Table 2, I categorize each sample member by the average number of weeks worked during the junior and
senior academic years 0, 1–10, 11–20, or 21 1 h per week, and summarize the high school transcript data for
respondents in each category. I confine my attention to the courses taken in grades 11 and 12, and I group the
courses into four aggregate categories: humanities and social studies, mathematics and science, vocational, and
all other courses. The five most frequently reported course titles in each category are listed in the note to
Table 2.
Table 2 reveals a striking difference in the curricular choices of the four “types” of high school students: those
who work the most intensively 21 1 h per week take fewer academic courses and far more vocational courses
than their counterparts who work moderately or not at all. Focusing first on the top panel of Table 2, we see
that high school students who work 0–20 h per week take about 3.8 Carnegie credit hours in the humanities and
social sciences, on average, while students who work over 20 h a week average only 3.5 Carnegie units. This
statistically significant difference in means of about 0.3 Carnegie units represents roughly 50 fewer hours in the
classroom over a 2-year period. Students in the 21 1 employment category also average fewer Carnegie units
296 A. Light Economics of Education Review 18 1999 291–309
Table 2 High school credits and grade point average in each subject area by average hours worked per week in high school
Average hours worked per week in grades 11–12 1–10
11–20 21 1
Subject area
a
Mean S.D.
Mean S.D.
Mean S.D.
Mean S.D.
Number of credits taken Humanities and social studies
3.76 1.20
3.86 1.39
3.78 1.62
3.49
b,c
1.30 Mathematics and natural science
0.94 0.88
0.94 0.95
0.81 0.85
0.77
c
0.80 Vocational subjects
3.26 2.09
3.37 2.23
3.94
b
2.25 4.10
c
2.10 Other subjects
2.09 1.47
2.03 1.46
1.88 1.34
1.46
b,c
1.37 All courses
10.04 1.89
10.20 1.99
10.41 1.99
9.82
b
2.03 Percentage of total credits taken
Humanities and social studies 38.07
12.47 38.17
12.26 37.48
13.16 35.93
12.12 Mathematics and natural science
9.32 8.44
9.30 9.33
7.67
b
8.11 7.78
c
7.84 Vocational subjects
32.09 19.14
32.53 20.13
37.02
b
19.47 41.66
b,c
19.54 Other subjects
20.52 13.37
20.00 13.98
17.82
b
12.04 14.62
b,c
13.17 Grade point average
Humanities and social studies 1.91
0.75 1.87
0.64 1.92
0.67 1.90
0.72 Mathematics and natural science
1.44 1.29
1.45 1.23
1.35 1.22
1.29
c
1.23 Vocational subjects
2.27 1.03
2.13 1.01
2.28 0.97
2.45 0.95
Other subjects 2.22
1.12 2.09
1.16 2.22
1.13 1.86
b,c
1.36 All courses
2.24 0.66
2.17 0.59
2.20 0.60
2.30 0.65
No. observations percentage of 685 143
20.9 245
35.8 170
24.8 127
18.5
a
The five most frequently reported course titles in each category, in descending order, are HumanitiesSocial Studies: American History, English III, American Government, English IV; Popular Literature; MathematicsScience: Chemistry I, Algebra II, Physics
I, Geometry I, Biology; Vocational: General Work Experience, Typewriting I, Accounting, Automobile Mechanics, Auto Shop I; Other: Physical Education, Health, BandOrchestra, Driver Education, Art I.
b
The null hypothesis that the difference between this mean and the one in the preceding column is zero is rejected at a 10 signifi- cance level.
c
The null hypothesis that the difference between this mean and the one in the left-most column is zero is rejected at a 10 signifi- cance level.
in mathematics and science courses than their less inten- sively employed counterparts, fewer credits in “other”
subjects, and fewer credits overall. At the same time, they tend to take significantly more courses in vocational
subjects such as “general work experience” including off-campus internship programs, typewriting, and auto-
mobile mechanics. The middle panel of Table 2 shows that students in the 21 1 category devote 42 of their
classroom time to vocational subjects, on average, which is far more than the other groups. As indicated by the
superscripts in Table 2, the null hypothesis that the mean number of credits for the 21 1 group is different than
the nonworkers’ mean is rejected at a 10 significance level for all four subject areas.
The bottom panel of Table 2 shows that students who work 21 1 h per week tend to receive lower grades than
their nonemployed counterparts in math, science, and “other” subjects, while performing slightly better in their
vocational courses. The average grade point average in math and science falls from 1.44 to 1.29 as we read
across Table 2, and the difference between the right-most and left-most means is statistically different than zero at
a 10 significance level. At the same time, the average grade point average in vocational subjects is almost 0.2
points higher for the most intensive workers than for the nonworkers. These statistics indicate that students tend
to receive the highest grades in the subject areas where they take the most courses, presumably because they
make curricular choices on the basis of aptitude. Interest- ingly, the grade point average for all courses taken in
grades 11 and 12 does not differ significantly among the four “types” of students: the mean, overall GPA is
between 2.2 and 2.3 for all four categories.
10
Table 3 extends Table 2 by showing how the four
10
Adding college-goers to the sample produces three changes in the summary statistics: 1 the mean number of credits in
humanitiessocial studies and mathscience is higher for all employment categories, 2 the grade point averages are higher
in each category, and 3 additional differences in means are statistically significant because of the larger sample sizes,
although the overall patterns are identical to what is seen in the reported version of Table 2.
297 A. Light Economics of Education Review 18 1999 291–309
Table 3 Characteristics of sample by average hours worked per week in high school
Average hours worked per week in grades 11–12 1–10
11–20 21 1
Mean S.D.
Mean S.D.
Mean S.D.
Mean S.D.
High school quality Student–teacher ratio
19.10 3.43
18.65 3.69
19.53
c
3.52 19.28
4.01 Average salary for first-year teachers
a
10.78 1.12
10.73 0.82
10.91
c
0.89 10.80
0.98 1 if school has distributive education program
0.59 0.56
0.65
c
0.72
d
Ability Score on academic components of ASVAB
60.76 23.40
66.76
c
22.95 73.10
c
22.70 71.77
d
20.46 Score on nonacademic components of ASVAB
89.19 30.63
99.58
c
30.07 107.95
c
30.97 110.59
d
27.92 Family background, demographics
1 if foreign born 0.06
0.03 0.06
c
0.04 Per capita of family income in grade 12
a
3.21 2.52
4.61
c
3.37 5.00
3.38 6.48
c,d
4.22 Number of siblings
4.84 3.15
3.99
c
2.28 3.59
c
2.15 3.83
d
2.99 Mother’s highest grade completed
9.88 2.85
10.68
c
2.42 11.10
c
2.27 10.96
d
2.44 1 if white
0.40 0.57
c
0.67
c
0.71
d
1 if black 0.43
0.32
c
0.24
c
0.17
d
1 if Hispanic 0.17
0.11 0.09
0.12 Market characteristics
Area unemployment rate in grade 12 9.02
3.54 9.19
3.67 8.66
c
3.28 9.07
3.95 1 if live in urban area in grade 12
0.71 0.66
0.69 0.71
Post-high school employment, wages Average hours worked per week
In first year after high school 18.59
15.10 19.51
c
13.62 27.44
c
14.33 33.61
c,d
12.86 In first 6 years after high school
28.87 12.52
26.60 12.42
31.77
c
12.74 37.93
c,d
10.47 Average hourly wage
a
1 year after high school 4.46
1.53 4.54
1.95 4.62
1.98 5.11
c,d
2.10 6 years after high school
5.34 3.26
5.91 5.42
6.12 2.80
7.64
c,d
5.86 After gaining 1 year of post-HS experience
b
4.34 1.76
4.63 2.39
4.58 1.91
5.07
c,d
2.13 After gaining 6 years of post-HS experience
b
6.04 4.02
6.32 3.23
5.86 2.42
6.94
c,d
3.25 Number of observations Percentage of 685
143 20.9
245 35.8
170 24.8
127 18.5
a
In 1982 dollars. All but average hourly wage are divided by 1000.
b
1800 h of cumulative, post-high school work experience is defined as 1 year.
c
The null hypothesis that the difference between this mean and the one in the preceding column is zero is rejected at a 10 signifi- cance level.
d
The null hypothsis that the difference between this mean and the one in the left-most column is zero is rejected at a 10 signifi- cance level.
“types” of individuals differ in terms of their high school, family, and personal characteristics, as well as in
their post-secondary employment and wages. The top rows, which summarize characteristics of the respon-
dents’ high schools, reveals that typical measures of high school quality student–teacher ratio and average teacher
salary do not differ systematically across the four categ- ories. Students who work intensively in high school are
neither more nor less likely than their less employed classmates to attend “good” schools. However, their high
schools are far more likely to have distributive education programs, which are designed to combine classroom
training with internships and other forms of “real world” experiences. Over 70 of the students who average more
than 20 h of work per week have access to such pro- grams, while fewer than 60 of the less intensively
employed students do so. This correlation is unsurpris- ing, for one would expect distributive education pro-
grams to facilitate the entry of high school students into the labor market.
Just as Table 2 indicates that high school employment and performance in vocational courses are positively
related, the next rows of Table 3 reveal that performance on the Armed Services Vocational Aptitude Battery
ASVAB increases systematically with high school work intensity.
11
I sum the raw scores for the general science, arithmetic reasoning, word knowledge, para-
graph comprehension, and mathematical knowledge tests
298 A. Light Economics of Education Review 18 1999 291–309
to obtain the “academic” score, while the “nonacademic” score is the sum of the raw scores for the numerical oper-
ations, coding speed, auto and shop information, mech- anical comprehension, and electronics information tests.
As Table 3 shows, mean scores on both the academic and nonacademic portions of the ASVAB increase sig-
nificantly as we move from the nonworkers to the stu- dents who average 1–10 h per week, and again as we
move from the 1–10 category to the 11–20 category. The upward trend then levels off, with the two most intensive
employment categories exhibiting mean scores that are statistically indistinguishable.
Among the family background, demographic, and market characteristics considered in Table 3, a number
of interesting contrasts emerge. First, there is a clear, positive relationship between high school employment
intensity and per capita family income: students who do not work in grades 11 and 12 come from families with
an average, annual income of US3210 per capita, while those who average over 20 h per week have an average,
annual, per capita family income of US6480. One explanation for this pattern is that the money earned on
jobs held in high school contributes to family income. A more likely explanation is that students who work tend
to come from families where one or more parents work continuously throughout the year. Students from such
families may learn about employment opportunities from their employed parents or even obtain jobs at their par-
ents’ work places. Second, students’ work efforts are also positively correlated with their mothers’ and,
although it is not reported in Table 3, fathers’ schooling attainment. Parental schooling is positively correlated
with parental employment level and earnings, so this pat- tern is consistent with the preceding one. Third, race and
high school employment are strongly related. Among the 143 males who do not work at all during grades 11 and
12, 40 are nonblack, non-Hispanic henceforth referred to as “white”, 43 are black, and the remaining 17
are Hispanic. Among individuals averaging more than 20 h of work per week, 71 are white and only 17
are black. This pattern is consistent with the well known finding that young, black men are more likely than non-
blacks to be nonemployed. Such patterns are generally attributed in part to the fact that blacks are concentrated
in economically depressed urban areas where jobs are scarce, but Table 3 shows that nonemployed high school
males do not differ significantly from others in their tendency to live in urban areas or in locales with above-
average unemployment rates.
The bottom rows of Table 3 reveal that high school employment intensity is positively related to post-high
11
The ASVAB was administered to virtually all NLSY respondents in the summer and fall of 1980, when the members
of my sample were still in high school or had just graduated.
school employment. The typical male who averages more than 20 h of work per week during high school
works an average of 34 h per week in the following year and 38 h per week in the following 6 years. This is sub-
stantially more work effort than is seen among the “inter- mediate” workers and especially the nonworkers, who
average only 19 h per week in the year after high school. Individuals who work 21 1 h a week in high school
also tend to earn higher wages than their less employed counterparts when a fixed amount of post-graduation
time has elapsed. One year after graduation, for example, their
average wage
is US5.11h,
versus about
US4.50h for the other groups. There are also statisti- cally significant, but smaller, differences in mean wages
among these “types” of students when the amount of actual, post-high school work experience as opposed to
time elapsed, or “potential” work experience is held constant.
3. Wage model