NORC at the University of Chicago
The Nonequivalence of High School Equivalents
Author(s): Stephen V. Cameron and James J. Heckman
Source: Journal of Labor Economics, Vol. 11, No. 1, Part 1: Essays in Honor of Jacob
Mincer (Jan., 1993), pp. 1-47
Published by: The University of Chicago Press on behalf of the Society of Labor
Economists and the NORC at the University of Chicago
Stable URL: http://www.jstor.org/stable/2535183
Accessed: 05-03-2018 14:34 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
NORC at the University of Chicago, Society of Labor Economists, The University of
Chicago Press are collaborating with JSTOR to digitize, preserve and extend access to Journal
of Labor Economics
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
The Nonequivalence of High
School Equivalents
Stephen V. Cameron, University of Chicago
James J. Heckman, University of Chicago
This article analyzes the causes and consequences of the growing proportion
of high-school-certified persons who achieve that status by
exam certification rather than through high school graduation. Examcertified
high school equivalents are statistically indistinguishable from
high school dropouts. Whatever differences are found among examcertified
equivalents, high school dropouts and high school graduates
are accounted for by their years of schooling completed. There is no
cheap substitute for schooling. The only payoff to exam certification
arises from its value in opening postsecondary schooling and training
opportunities, but completion rates for exam-certified graduates are
much lower in these activities than they are for ordinary graduates.
This article examines the causes and consequences of a neglected social
phenomenon-the recent rapid growth in the fraction of persons who
achieve high school certification by means of an equivalency exam rather
than through the traditional route of high school graduation. In 1968, only
5% of all new high school certificates were awarded through equivalency
This research was sponsored by National Science Foundation grant SES-91-
114551 and a contract from the Bureau of Labor Statistics, U.S. Department of
Labor. The revision was sponsored by a grant from the Lynde and Harry Bradley
Foundation of Milwaukee, Wisconsin. Janet Baldwin, Ricardo Barros, Gary Becker,
Ken McLaughlin, Paul Siegel, Bob Topel, and Joe Tracy made valuable comments
on versions of this article. Joe Hotz and Seth Sanders kindly gave us their data on
local labor markets that supplement the National Longitudinal Survey of Youth
Geocode Data. Seymour Brandwein, Georgia Goeters, and Chris Taber provided
excellent research assistance.
[Journal of Labor Economics, 1993, vol. 11, no. 1, pt. I]
? 1993 by The University of Chicago. All rights reserved.
0734-306X/93/ 101-0010$01.50
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
2 Cameron/Heckman
exams. By 1987, the corresponding figure was in excess of 14%. In 1968,
only 2% of all persons who completed their education with high school
degrees were exam-certified. In 1987, the corresponding figure was almost
11/.
It is widely believed that exam-certified high school equivalents are
equals of traditional high school graduates in all relevant behavioral dimensions.
This view is fostered in part by the American Council on Education,
a private organization representing institutions of higher education
as well as regional education associations. That organization administers
the most widely used equivalency exam, the GED (for general equivalency
diploma). Researchers affiliated with the American Council on Education
claim that "persons who meet state/provincial established minimum score
levels for the high school equivalency credential based on GED tests should
be considered high school graduates for admissions, military, licensing and
employment purposes. The test results ... demonstrate this achievement
equivalency" (Malizio and Whitney 1982, p. 10). The two types of high
school certification are treated as identical in the U.S. Census and the
Current Population Surveys (CPS) widely used to monitor national social
and economic well-being.
This article challenges the conventional wisdom. Exam-certified high
school equivalents are not identical to traditional high school graduates in
terms of their ability as measured by a standard psychometric test (the
Armed Forces Qualifying Test), in terms of their wages and hours of work
or in terms of their postcertification education and training decisions. Examcertified
high school equivalents are psychometrically inferior to traditional
high school graduates. Elsewhere we establish (Cameron and Heckman
1991, 1992) that the determinants of high school certification by exam are
very different from the determinants of traditional high school graduation.
We demonstrate here that the economic consequences of the two avenues
of high school certification are quite different. Exam-certified persons are
indistinguishable in many relevant labor market dimensions from high
school dropouts who are uncertified. Differences in wages among high
school graduates, exam-certified equivalents, and dropouts are accounted
for by years of schooling attained. There is little evidence of value added
from exam certification beyond the effect of years of schooling completed
on wages. However, exam-certified graduates are more likely to take vocational
and technical training, while traditional high school graduates are
more likely to attend academic 4-year colleges and complete academic
programs when they begin them. Exam-certified high school graduates are
more likely to participate in some form of postsecondary training than are
non-exam-certified high school dropouts. Exam-certified persons who take
postsecondary schooling and training earn lower wages than high school
graduates undertaking the same activity. The return to high school equivalency
certification, such as it is, comes from the return to postsecondary
training. Accordingly, it is not appropriate to consider the GED as an
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 3
educational end in itself-the emphasis in many contemporary state and
federal programs.
In light of our evidence, the growing use of GED certification suggests
the possibility of widespread misperception on the part of test takers. We
establish in this article that the growth in the level and proportion of examcertified
high school credentials is a direct consequence of federal and state
human resource policies that make GED test taking privately rational even
if it is socially unproductive. Since the rnid-1960s, both federal and state
governments have increasingly subsidized adult basic education programs
that have placed a growing emphasis on adult equivalency as a clearly
identified and desirable objective. In addition, a high school degree or an
exam-certified equivalent is required for participation in a host of postsecondary
vocational and academic financial support programs increasingly
subsidized by federal and state governments over this period. The demand
for participation in these subsidized programs induced a derived demand
for high school certification on the part of high school dropouts.
One major conclusion of this article is that the GED is a vehicle for
participation in postsecondary education due to its value in satisfying bureaucratically
determined qualifications for admission and financial support.
Subsidization of these programs by governments reconciles the apparent
conflict between low economic returns to obtaining the GED and the large
and growing demand for GEDs.
A second major conclusion concerns the limited value of psychometric
measurements for predicting labor market outcomes. Our evidence about
the labor market inadequacy of exam-certified graduates calls into question
the value of psychometric evidence on the efficacy of private schools. (See
Coleman, Hoffer, and Kilgore [1982]; and Chubb and Moe [1990], who
rely on such evidence.) Tests like the GED measure skills that are only
weakly related to the skills valued by employers.
This article develops in the following way. Section I documents basic
facts about high school equivalency. Psychometric and market evidence
demonstrate the nonequivalence of high school equivalents. Section II presents
more refined evidence on the economic returns to high school equivalency.
Section III presents reasons for growth in exam certification in the
presence of low economic returns from the activity. The article concludes
with a summary.
I. The Changing Structure of High School Certification and
Its Consequences
A. The Growth in High School Equivalency Certification
There are three main routes through which Americans achieve certification
as high school graduates: (a) through traditional course attendance,
culminating in graduation at the end of the twelfth grade; (b) through
night school and other formal schooling programs for those who drop out
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
4 Cameron/Heckman
of traditional high school programs; and (c) through certification on a
standardized exam for high school dropouts. Although the vast majority
(84.5% in 1987) of all new high school credentials are issued through
traditional route a, a sizable proportion of new graduates come from the
less traditional avenues b and c. The largest nontraditional source is from
persons certified by an equivalency exam-roughly 14% of all newly issued
high school credentials obtained in 1987 were secured by this means. Virtually
all of these credentials come from individuals who passed the nationally
normed GED exam developed by the American Council on Education.
Graduation through formal adult secondary schooling produced
no more than 2% of all new high-school-certified persons in 1987.
There has been a dramatic change in the number of exam-certified high
school graduates over the period 1953-88. Figure 1 plots the percentage
of GED recipients relative to all high school graduates for each year over
the period. (The reasons for this growth, summarized below the figure,
are discussed in Sec. III below.) It rises from less than 2% in 1954 to more
than 14% in 1986. The period 1965-85 is one of especially rapid growth.
There has been concomitant growth in the percentage of all persons with
high school diplomas (and no further academic degree) who achieve that
status by GED certification. Figure 2 reveals that, of the total stock of
persons with only high school degrees by 1987, more than 10% achieved
their degree by taking a GED exam. In 1968, only 2% of the total stock
was exam-certified. Figure 3 documents the near stability in non-GED
sources of nontraditional high school graduates. Certification through adult
education courses ("other programs") has grown over the period 1974-
87, but the level is low (ranging between /% and 2% of all new high school
graduates), and the growth rate is small. The major change in the source
of high school credentials is growth in GED certification.
The GED testing program began in 1942 as the Veterans Testing Service
and was a joint venture of the United States Armed Forces Institute and
the American Council on Education. One premise of the testing program
was that the life experience of military personnel could substitute for classroom
training in developing skills associated with high school certification.
A second premise was that the relevant skills could be measured by an
exam. By 1952, all but three states issued certificates of high school equivalence
to veterans and servicemen who passed the Veterans Testing Service
exam. The armed forces then accepted exam-certified equivalents as the
equals of high school graduates in making their enlistment and screening
decisions-even for service academies. A commission on accreditation of
service experiences in 1952 documented the widespread acceptance of the
GED as a high school certificate by major firms and state and local governments.
In that same year the American Council on Education began to
offer the exam to nonveteran civilians, and its name was changed to the
GED. By 1963, all 50 states used the GED exam to certify high school
dropouts.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 5
B. Some Features of the Recent GED Exam and
Those Who Take It
The age distribution of GED test takers has remained roughly constant
over time, although the influence of the baby boom and subsequent baby
bust on the time series of exam-certification rates is evident. Most GED
test takers are less than 25 years old. Assuming temporal stability of pass
rates by age, the baby boom accounts for part of the post-1970 growth in
GED-certified graduates as a fraction of total high-school-certified persons.
Between 1970 and 1987, the ratio of 16-19-year-olds to 20-24-year-olds
fell from .89 to .75. Over the same period, the proportion of persons age
17 relative to ages 20-44 declined from .056 to .040. Relatively more persons
were in the age brackets at risk for the GED than in the age brackets at
risk for traditional high school graduation. However, rough calculations
suggest that changing population proportions by age can account for, at
most, 2 points of the 8-percentage-point growth in GED-certified persons
as a proportion of total new certified persons that occurred over this period,
and it does not explain the increase in test taking over the whole period.
The growth in exam-certified equivalents explains an apparent contradiction
in the data on high school dropouts. Figure 4 plots the proportion
of traditional high school graduates for cohorts of 17-year-olds over the
period 1951-88. The proportion declines after 1968, although it slightly
rebounds in the late 1970s and 1980s. The figure shows a very different
pattern over the period 1971-86 for high-school-certified persons ages 20-
24. The recent growth in exam certification explains the discrepancy between
the two figures (see Finn 1987). There appear to be sharp differences
in the use of GED certification by race. Table 1 documents that black
CPS-measured high school equivalents are almost twice as likely as whites
to possess a GED. Part of the measured convergence of black and white
high school attainment rates demonstrated by Kominski (1990) is due to
the growing high school certification of blacks by GED examination.
High school certificates awarded by adult education institutes reward
students for completing a traditional high school curriculum at a somewhat
later stage of life than do typical high school graduates. Equivalency examination
programs operate on a radically different principle. Since the
GED certifies the vast majority (well in excess of 90%) of all exam-certified
high school graduates over the period 1970-87, we focus our attention on
that exam.
GED candidates are tested on a total of 290 items in five subject area
tests: writing skills (80 items), social studies (60 items), science (60 items),
reading skills (40 items), and mathematics (50 items). Conceptual-and
not factual-knowledge is stressed. The focus is on general knowledge
and not specific details (Malizio and Whitney 1982). Individual states set
pass standards, but these vary within a fairly narrow band. The majority
of the states (29) require a minimum score of 35 (out of 80 possible-20
is the minimum score) on each exam and an average of 45 over all exams.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
(D
CD
co
0
lot
u
0
LO
t0
o =0
N
;--
00
CD~ I Cs0 C C
a,~~ o
6~ oT
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
A NA"
G.~~~~~~~~~~O
_ ' ? .?o0
CZ
o 0 w 3
.0 = - 0
0 o _ _
0 *' o
00~~~~~~~~E
oAL
oo
L O*
t 0
ff . .5= c sc
- z .s .
o - 0 C/ Q
7
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
8 Cam eron/Heckmran
12 -
10
4-
2-
0
54 58 62 66 70 74 78 82 86 90
Year
FIG. 2.-GED recipients as a percentage of the total stock of persons ages 15 and older
with high school credentials only (GED + high school graduates only). Source: see fig. 1.
Most of the rest require a minimum of 40 and an average of 45 (GED
Testing Service 1989, p. 30). Graduating high school seniors are used to
norm the test. Fifty percent of graduating high school seniors score 50
points or higher.
Because the minimum pass level is set at 35 on the distribution of graduating
seniors with a range of scores set at 20-80, GED graduates necessarily
outperform graduating high school seniors on the test over a range of test
scores.' This is an artifact of test construction, although Malizio and Whitney
(1982) offer such evidence to conclude that GED-certified persons are
the equals or superiors of high school graduates. Below and elsewhere
(Cameron and Heckman 1992), we demonstrate that GED recipients are
psychometrically inferior to high school graduates in terms of their performance
on the Armed Services Qualifying Test and its components.
Candidate preparation for the GED is limited. In April and May 1980,
a survey was conducted of 13,000 GED candidates at 250 randomly selected
GED testing centers throughout the United States. The median examinee
spent 20 hours preparing for the test and spent $10 in preparation costs.
Seventy-five percent of the examinees spent 60 hours or less in preparation.
The upper fifth percentile reported more than 200 hours in preparation.
The upper quartile of the candidates spent $25 in direct-out-of-pocket
costs or $30, including lost salary (Malizio and Whitney 1981, table 18).
More precisely, GED recipients must score better than high school graduates
scoring less than 35. If the distributions of ability were identical in the two groups,
then mean, median, and all quantile test scores would necessarily be higher for
GED recipients. The evidence in Malizio and Whitney (1982) suggests that GED
recipients have a thinner right tail of test scores compared to high school graduates,
so their mean score could still be lower than that of high school graduates, but in
fact it is higher.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 9
Percent
0.20-
0.15 -
0.10 -
0.05 -
0.00
e8 70 72 74 76 78 80 82 84 86 88 90
Year
Private School -------- Other Programs ------ GED Recipients
Percent
0.90
0.85
0.80-
0.75 -
0.70 -
68 70 72 74 76 78 80 82 84 88 88 90
Year
Public School
FIG. 3.-Proportion of individuals completing high school by type of program. Source:
U.S. Department of Education (1989); GED Testing Service (1990).
Even at the upper fifth percentile point in the distribution of costs, the
corresponding figures are $100 and $106. Twenty-one percent did not
prepare in any way. Only 22% took the GED practice test, and 40.5%
studied from a book or manual. Less than 1% of the candidates incurred
any expenses for individual tutoring. Despite the generally low level of
preparation, usually more than 70% of those taking the exam pass it in
any given sitting. Candidates who fail may retake the exam without penalty,
although there is a short (2-3 month) waiting period in some states. In
1991 we observed a federally sponsored GED program that gave persons
initially certified at fourth-grade levels in numeracy and literacy 4 weeks
of intensive instruction. The program has a first-time pass rate of 80%. If
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
10 Cameron/Heckman
90
P 80
e 7
570 -
t
60 -
50 55 60 65 70 75 80 85 90
Year
Age 17 HS Grads - - 20-22 wI Dlploma
FIG. 4.-The percentage of 17-year-olds who are high school graduates and the percentage
of 20-24-year-olds with at least a high school diploma. The 17-year-old graduates include
graduates of regular day programs and excludes graduates of other programs, when separately
reported, and high school equivalency recipients. Source: U.S. Department of Education
( 1989); U.S. Bureau of the Census (various years).
GED certification is so easy to attain, it is natural to conjecture that its
intrinsic economic value might be low.
C. Psychometric and Other Evidence on the Nonequivalence
of Exam-certified Equivalents
There is considerable evidence that GED-certified persons do not possess
the same skills or motivation as high school graduates. Laurence (1983,
table 1) notes that high school dropouts and GED-certified high school
equivalents had basically the same attrition rates from the U.S. military
over the period 1977-79, and both groups attrited at twice the rate of high
school graduates. She goes on to note that, in 1982, the U.S. Army required
for minimal admission standards that GED-certified graduates and high
school dropouts should be in the thirty-first percentile of the Armed Forces
Qualifying Test (AFQT) distribution. High school graduates were only
required to be in the sixteenth percentile. The higher minimum scores
were judged necessary to guarantee successful completion of basic training
courses by GED-certified applicants. Recently (1991), the U.S. Army has
refused to accept persons possessing only a GED.
A recent study of Iowa GED test takers (Beder 1992) claims strong
positive effects of GED certification on labor market outcomes. The evidence
in that study is based on before-after comparisons for those who
attain the GED. Contrary to accepted practice, the study uses no control
group of high school dropouts or high school graduates, as we do below.
It attributes all of the labor market effects of life-cycle work experience,
job changing, maturation, and geographic mobility to the GED. It greatly
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 11
Table I
Proportion of Individuals Receiving a GED or Graduating High School
Black Hispanic White
A. Age 25 males:
GED .11 (.01) .11 (.01) .07 (.01)
High school graduate .68 (.02) .60 (.02) .81 (.01)
% of Current Population
Survey measured high school
equivalents who are GED
certified 14 15 8
Sample size 1,088 693 1,820
B. Age 25 females:
GED .07 (.01) .11 (.01) .07 (.01)
High school graduate .77 (.01) .64 (.02) .84 (.01)
% of Current Population
Survey measured high school
equivalents who are GED
certified 9 13 8
Sample size 1,082 674 1,719
NOTE.-These proportions do not chan
87 waves of the NLSY (see App. A). I
above calculations. Standard errors of
exaggerates the contributio
provements.
An extensive study of the
versity of Wisconsin system was performed by Pawasarat and Quinn
(1986). At the University of Wisconsin-Milwaukee, GED-certified persons
had lower completion rates for the first four semesters of college
attendance than did high school graduates from the bottom twentieth percentiles
of their high school class (310% vs. 41j%). The 4-semester-completion
rates for high school graduates in the top 50% of their class was 62%twice
that of the GED graduates. At the University of Wisconsin-Madison,
only 73% of GED holders admitted to the school enrolled for a second
semester, compared to a 95% rate for all entrants to the school. At Milwaukee
Area Technical College, a vocational school, GED holders seeking
a 2-year associate degree had attrition rates comparable to those of high
school dropouts. Over the period 1980-83, 8% of GED entrants attained
the 2-year associate degree, compared to 10% of high school dropouts and
30% of high school graduates. A study of the Milwaukee labor market by
Pawasarat and Quinn finds that 48% of the firms interviewed preferred
hiring conventional high school graduates to GED-certified graduates,
while the rest of the employers were indifferent between persons wich the
two types of high school credentials.
Psychometric evidence from the National Longitudinal Survey of ' oi- h
(NLSY), ages 13-20 in 1978, contradicts the claims of psychometric
equivalency for the two types of high school certification made by the
American Council on Education. (The survey is described in App. A.)
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
12 Cameron/Heckman
Table 2 displays the resul
administered to all members of the NLSY sample. For the random subsample
of the data, AFQT scores for the deciles going from the bottom
to the top are presented, as well as mean scores. High school graduates
have statistically significantly higher mean test scores than do GED holders,
who, in turn, have statistically significantly higher mean test scores than
do high school dropouts. The pattern is the same at each decile of the test
score distribution from the top to the bottom. The pattern holds true for
a sample standardized to have the same approximate age, 16 or 17, at the
time they take the exam, or for an enriched sample that oversamples blacks
and Hispanics (results are not shown). The same pattern is found for
persons who do not complete 4-year colleges both for the entire sample
and for those who were 16 or 17 at the time the test was given: GED
recipients are not the test-taking equivalents of high school graduates.
(Cameron and Heckman 1992). However, they dominate high school
dropouts in test taking. Similar patterns appear for each race group: whites,
blacks, and Hispanics. A Wilcoxon test for stochastic dominance (Bickel
and Doksum 1977)-a statistical concept that compares distributions of
the same outcome for different groups and determines if higher outcomes
are more common in one group than another-is presented in the first
row of table 3. It reveals that the high school graduate AFQT distribution
first-order-stochastically dominates the GED-AFQT distribution, and the
latter distribution first-order-stochastically dominates the AFQT distribution
of high school dropouts. The same pattern is found for persons
who do not attend college and for disaggregated components of the AFQT
exam (see Cameron and Heckman 1992).
D. Direct Behavioral Comparisons
This subsection presents simple mean-difference and univariate distributional
comparisons among high school dropouts, GED recipients, and
high school graduates. Using the NLSY data for male youth ages 13-20
in 1978, we compare the determinants and labor market and educational
consequences of the three types of high school certification status.
Table 4 reveals that high school dropouts are more likely to be minority
group members and come from larger families with lower incomes and
less educated parents than do GED recipients, who, in turn, have poorer
background characteristics than high school graduates.3 A Wilcoxon test,
2 These results are available on request from us in a separate App. B. This appendix
was not included here for the sake of brevity.
3 The anomalously high number of siblings is a consequence of size-biased sampling
in the NLSY. If one child is included in a unit, so are all of his or her
siblings provided they share common family characteristics. This sampling induces
a stochastic dependence among sibling observations that we analyze elsewhere
(Cameron and Heckman 1992), where it is shown to have a minor effect on the
estimated standard errors of the coefficients of wage equations.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 13
Table 2
Means and Deciles of Test Scores on the AFQT Exam
for the Random Sample
Deciles (Lowest to Highest)
Mean
N (SE) 10 20 30 40 50 60 70 80 90
High school graduate 2,168 75.8 (0.40) 48 61 68 74 79 84 88 93 97
GED 209 64.7 (1.28) 38 48 54 61 66 70.5 76 82 88.5
Dropout 436 45.5 (0.79) 25 30 35 39 43 48 53 60 70
Total 2,813 70.1 (0.40) 37 49 60 68 74 80 85 90 96
NOTEI.-The sample was constructed LISingI the 1979portion
of the data was uLsed above. Approximately 6
similar Lising the combined black, Hispanic, and randomn
reported in table 3, reveals that the fam
high school graduates stochastically dominates (i.e., has more weight on
higher incomes) that of GED recipients and dropouts. The family income
distribution of GED recipients in turn dominates the family-income distribution
of dropouts.
Table 5 presents evidence on labor market outcomes for individuals
with the three types of high school degree status. At age 25 (table 5, pt.
A), the mean labor market status of high school dropouts is the same as
that of GED recipients. The small premium in hourly wages and salary
for GED recipients over that of dropouts is not statistically significant.
Both groups are inferior to high school graduates in terms of hours, wages,
salaries, weeks worked, and length of time on their current job. The lower
work experience of high school graduates is a consequence of their greater
Table 3
Wilcoxon Tests for First-Order Stochastic Dominance (with x2
Approximations): Family Background and Outcome Measures,
Random Samples
High School > GED GED > Dropout
Variable X2 Prob > x2 X2 Prob > x2
A. 25-year-olds:
AFQT score 46.04 .0001 68.40 .001
Annual hours 7.31 .0068 .04 .840
Hourly wage 10.31 .0001 2.37 .130
Family income 28.30 .0001 11.40 .007
County average earnings
(unskilled) 4.31 .0381 3.17 .075
B. 28-year-olds:
AFQT score 20.9 .0001 20.6 .0001
Annual hours 4.84 .028 1.02 .31
Hourly wage 9.71 .0014 .78 .38
Family income 13.61 .0002 2.26 .133
County average earnings
(unskilled) .82 .52 .10 .77
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
o m t- i--
o o
C) CO OCO
00s .2
COz-00r C, r:'-C:)c
GoI'2.~NND -t
Cr .E- 0-CrCo0o (N
02 V1 /
O~~~~~I CIO 'O
oo~~~ rIONO ~~~CI
,c l
Z o 0.C22 Go
S 0) b
~~J*) 0C C 0
14
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
3 C. ? Ln > < ) .
u~~~~~~~~~~~
CIA 4 7 s 00 u D.
CIA
00 s 00
bXSn7,aN>D-?v
0 4 C --' -4--
0 \ i r i-- O co D
.-0 " 06 _: C>_0C-i
? - 00 CI - 0 * iJ
_ r0 1N Ln 5
toC)r\l,-,-Ntc.bJ~~~~s|1~~~ntuDl,t5
Cl~~~~~~~~~~E
0 C) C) Lr
XC., 4-1 . : l (-\I N i
43~~~~~~~~~~~C
C-I 7 l
O vl LK itif-) ?v
c41)S_:=n3~~~~~~~~- c o o L ) L.
00~~~~
X 4-- Lr C) D
_~~~~~' ? 3r| Nt-N?1 On
-u 0 C
00 0 *c Kn :,,,s-A
t0 C) rD-I O 10 5
.0
ct CL , \ 0 5;. ) 5 t'- Cl 0
"-Q CC ,l 'o 5
U', -- ' - - u z
U4 C) *N-C*
15 ' - C) C)
U4 ~ ~ ~ ~ ~ U U
0~~~~~~~~~~
0 0 ~~~~~~
- Cl 0N C)~ 00
0 5
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
16 Cameron/Heckmian
schooling. The relationships in means carry over to first-order-stochastic
dominance on these variables: GED and high school dropouts are indistinguishable,
and both groups have labor market outcome distributions
that are first-order-stochastically dominated by those of high school graduates
(see table 3, pt. A, for 25-year-olds and pt. B for 28-year-olds).
Similar descriptions apply to table 5, part B.
A focus on mean outcomes, while traditional, masks much important
detail in the wage distributions. Figure 5a places GED recipients and high
school dropouts in the wage distribution of high school graduates at age
25. These figures plot the proportion of the dropout and GED wage distribution
located in the deciles of the high school graduate wage distribution.
A flat line at 10% in these figures would indicate that GED recipients
and high school graduates are indistinguishable from high school
graduates at all deciles of the high school graduate wage distribution.
Shortfalls or excesses around the 10% line indicate shortfalls or excesses
in the wage distribution of GED recipients or high school dropouts compared
to that of high school graduates. The near parity in the top 10% of
the GED distribution at age 25 might be termed the "Bill Cosby" effectnamed
for the most famous GED recipient. There is a group of highly
motivated and able GED recipients who achieve the goals of the program.4
For most GED recipients, however, disparity and not parity is the rule.
The mass of the GED recipients is located in the bottom half of the high
school graduate wage distribution. Placing GED recipients in the wage
distribution of high school dropouts (fig. 5b) demonstrates that, except
for the motivated top tenth percentile and the evacuated bottom tenth
percentile, there is a lot of similarity between the two distributions at
age 25.5
One way to gauge the economic significance of these results is to examine
their implications for the estimated "rate of return" to education arising
from the Current Population Survey convention that equates GED recipients
and high school graduates. Using a sample of NLSY observations of
young men age 25 (enriching the random sample with black and Hispanic
subsamples), we compute a least-squares regression of log hourly wages
on dummy variables that measure whether or not a person has a high
school diploma (- 1 if a person has a high school diploma irrespective of
subsequent achievement), or 2 years of college, or 4 years of college; see
4The median of the adjusted AFQT test for GED recipients in the top decile of
the high school wage distribution for those who do not go on to college is virtually
the same as that of the high school graduates. High school graduates in that decile
are at the sixty-eighth percentile in the AFQT distribution, while GED recipients
are at the sixty-third percentile in that distribution. High school dropouts at the
top decile of the high school wage distribution have a median AFQT score of 31.
5 The figures for 28-year-olds are strikingly similar. They are available on request
in App. B.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 17
Frequency
251
20 -
15
10
10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100
- Drop Out - l- GED Recipient FIG.
5a.-NLSY males, age 25: how the GED recipient and dropout hourly wage distributions
fit into the high school graduate wage distribution, by %. Source: NLSY (1979-87
data).
table 6, column 1. (Results for 28-year-olds are available in App. B, which
is available on request.) Column 2 shows the effect of distinguishing how
the high school diploma was achieved: through a GED or through a traditional
degree program.
Defining a high school diploma in the CPS-Census manner produces a
differential effect on wages at age 25 of 4-year college attendance compared
to high school graduation of 21% (col. 1). Breaking out the GED from
the traditional high school diploma produces a college/high school differential
of only 19.6% for the traditional high school degree. The comparable
figures at age 28 are 21.9% and 20.7%, respectively.6 F-tests based
on robust McKinnon-White (1985) jackknifed standard errors reject the
hypothesis that GED recipients should be considered the same as high
school graduates, but they do not reject the hypothesis that GED recipients
are indistinguishable from high school dropouts. The CPS-Census convention
of equating GED recipients to high school graduates overstates
the returns to college education relative to traditional high school graduation.
Inappropriate pooling of the two types of high school credentials
would cause the college/high school differential to increase over the period
of the late 1970s and 1980s, as GED certification became a widespread
phenomenon, but the effect appears to be relatively small. Approximately
3.6% of the growth of the 4-year college/high school differential documented
for younger workers (with 5 years of work experience) documented
by Katz and Revenga (1989) arises from falsely attributing the market
productivity of traditional high school graduates to GED recipients.
Additional evidence on the nonequivalence of high school equivalents
is presented in tables 7 and 8, which look at postcertification educational
6 The table for age 28 is available on request in App. B.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
1 8 Cameron/Heckmnan
Frequency
251
20
15 1
10
0
10 20 30 40 50 60 70 80 90 100
| GED Recipient I
FIG. 5b.-NLSY males, age 25: how the GED recipient hourly wage distribution fits into
the dropout wage distribution, by %. Source: NLSY ( 1979-87 data). See table 9 notes for a
definition of the sample.
choices for both types of degrees. Table 7 shows first choices after completing
certification. GED-certified persons are much less likely to attend
4-year colleges and are more likely to enter the military or not undertake
any postsecondary education. Table 8 reveals that GED graduates are less
likely than high school graduates to attend 2- or 4-year colleges or to
graduate from them if they attend them.
The evidence from the NLSY and the other studies indicates that GED
recipients are not the equivalents of high school graduates. Their labor
market outcomes and performance in the military suggest that GED recipients
are similar to high school dropouts. GED recipients are less likely
to pursue postsecondary academic education and are less likely to finish
an educational program they begin. Evidence from the Panel Survey of
Income Dynamics reported in Cameron and Heckman ('1992) corroborates
these results for age-groups comparable to those in the NLSY. In the
balance of this paper and in our companion papers (Cameron and Heckman
1991, 1992) we present a more refined statistical analysis of the NLSY that
supports these basic conclusions.
II. Econometric Evidence on the Nonequivalence
of Exam-certified Equivalents
A. Introduction
Unadjusted comparisons of means or distributions can be misleading.
This is likely to be true in evaluating the impact of the GED on labor
market outcomes. GED recipients attain 1 more year of high school before
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 19
Table 6
Log Hourly Wage Equations at Age 25 (N= 2,308)
Combine GED Disaggregate GED
and High School and High School
Graduate Graduate
(1) (2)
Intercept .690(22.9) .690(23.0)
High school graduate .130(5.5) .144(6.5)
GED N.A. .060(1.5)
2 years college + high school .231(5.0) .236(5.4)
2 years college + GED N.A. .169(1.1)
4 years college .340(10.2) .340(10.2)
Black - .190(9.1) - .190(9.1)
Hispanic -.050(1.8) -.050(1.8)
Year 1982 .024(.9) .023(.7)
Year 1983 .001(.0) .001(.1)
Year 1984 -.036(1.1) -.032(1.0)
Year 1985 -.071(1.1) -.070(9)
Year 1986* -.036(2.0) -.038(1.8)
F-test: probability > F:
GED = 0 N.A. .13
GED = high school
graduate N.A. .02
NOTE.-Persons in college at age 25 are deleted, as are persons not working. Wage equations are with
education dummies; all education dummies are defined exclusively. t-statistics are in parentheses and are
calculated using modified McKinnon-White standard errors to correct for heteroscedasticity. N.A. = not
applicable.
* Year 1987 is the left-out year indicator.
they drop out of school than do dropouts who do not attain the GED.
Unadjusted comparisons of GED recipients and dropouts would show
greater earnings for GED recipients, but this could be attributable solely
to attained years of schooling. GED recipients are older than high school
graduates and have more work experience. Both of these factors produce
a bias in favor of higher wages and earnings for GED recipients compared
to high school dropouts who do not attain the GED. These factors suggest
that the weak effects of the GED on economic returns demonstrated in
Section I are, if anything, overstated.
Table 7
Random Sample NLSY: First Action after Completing Degree (in %)
Attend Attend On-the-Job
4-Year 2-Year Vocational Training/
College College Training Apprenticeship Military Other*
Graduate high school
(N= 1,902) 36.6 23.3 7.6 3.4 4.8 24.3
GED (N = 164) 15.0 23 10 3.3 10 40
SOURCE.-NLSY (see App. A and the note
* Other = work with no training, unempl
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
20 Cameron/Heckm an
Table 8
Random Sample NLSY (in %)
A. All Educational Decisions after Receiving Degree
Attend 4-Year Attend 2-Year
College College No College
High school diploma
(N= 1,902) 30.3 32.3 37.4
GED(N=164) 16 27 58
B. Completion Rates for 4-Year College for Those Who Decide to Attend
4-Year College to Graduation*
Attend 4-Year Attend 2-Year
College College No College
High school diploma
(N= 566) 75 N.A. N.A.
GED (N = 42) 5 N.A. N.A.
C. Completion Rates for 2-Year College Attendance
Attend 4 Years Complete Less
and Graduate Finish 2 Years Than 2 Years
High school diploma
(N= 584) 34.7 21 44.3
GED (N = 42) 2 25 73
NOTE.-N.A. = not applicable. Only 3.2% of the sample attended a 2-year college and then went on
to a 4-year school.
* These are persons who start at 4-year colleges.
Conversely, if receipt of a GED starts a high school dropout on the
career path of a high school graduate, work experience at the new educational
certification level should be distinguished from that at the old in
comparing high school graduates and GED recipients. The latter persons
will have less work experience at the high school equivalent level than the
former. Failure to control for this difference biases downward the estimated
economic returns to exam certification if the return to work experience is
greater for high school graduates and GED recipients than for dropouts.
Although it is currently fashionable to ignore selection bias in studies
of labor market outcomes, it is likely to be an important problem in this
study. Wages are not available for nonworkers. Young persons may be
nonworkers because they are attending school or because they cannot get
a job. Persons with missing wages are unlikely to be typical of those for
whom wage data are available.
In this section, we examine the robustness of the evidence reported in
Section I to a variety of statistical adjustments. The simple story of that
section holds up in a more rigorous econometric analysis. Before turning
to the data, we first sound a cautionary methodological note.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 21
B. Sample Size and the Choice of a Significance Level
The evidence presented in this section of the article is largely based on
classical testing theory for multivariate regression models. Because we use
"robust" jackknife procedures (Efron 1982; or McKinnon and White 1985),
we do not rely on standard, and controversial, normality assumptions for
producing standard errors. Nonetheless, there is ambiguity in the classical
theory about the choice of a correct significance level for conducting tests
and how it should be adjusted in different sample sizes (Lindley 1957).
These considerations are especially relevant for this article in light of the
small samples available in the NLSY compared to those in the widely used
Current Population Survey.
In order to avoid placing undue-and increasing-weight on minimizing
type II errors (the probability of accepting a false null hypothesis) as
sample sizes increase, the probability of type I errors (i.e., the significance
level) should be adjusted downward with sample size. Given that p values
are to be used in judging hypotheses, we should be more tolerant-less
likely to reject a null for any p value-in a small sample like the NLSY
than in a large sample like the CPS or Decennial Census Microdata samples.
Two principles are important to keep in mind in reading the evidence
reported below and comparing it to evidence obtained from CPS or Census
surveys: (a) when one rejects a null hypothesis in a model fit on the NLSY,
one can be relatively confident in doing so; (b) when one does not reject
a hypothesis, but the sign pattern of estimated differences is plausible and
points to rejection, one should not be too confident in accepting a null
hypothesis of no difference.
C. The Direct Effects of Certification on Wages
and Hours Worked
We demonstrate that GED-certified males are more like high school
dropouts than high school graduates in terms of their labor supply and
wages. Table 9 presents estimates of alternative specifications of labor supply
and wage equations that distinguish GED recipients from traditional
high school graduates. We estimate wage and labor supply equations at
ages 25 and 28 for samples of young men not in college (2-year or 4-year)
at ages 25 or 28 who also are working at those ages. A second specification
reported in Cameron and Heckman (1992) is fit on samples of young men
who have not attended any college up to age 25 or 28 and who work in
the year following the date at which the age is attained. The evidence from
those samples corroborates the evidence reported here. The samples are
defined so that data on hourly wages are available for each observation
and so that persons holding low-wage part-time student jobs are excluded
from our analysis. In order to correct for potential sample-selection bias
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
x x ;e ~~~~O o C) * G *o rn loC
C
bs~~~~~C C) - CC C -o "- w - C) C)
o n X X t t . . . .o . t .
X tt ~~~~oo o - o mmo*n: C tiQ i 1-C).
0Q . . . . -0 hrALnC co -0c
0jC (0 IIC )r lArAC ) C
ogo 0 I I I
-e C m "
- 00 0C VQ l ,3 ,<} ,C
W~~~~~~~~~~r 0 Ol -!?t rl rl oo O L O N (D e
sZ E ; t . . . . 1- . 1- .
nn
5<4C;IIOO0S 0S 0L
0) '-.--0 C) * I ~ I C I nf )
m Cf 't N C D C) C)-00
"- = .o . . . .o . . . . . N
73 _ ._. *. . . .
0)Cn E -C 00 C) C) . . . .
U C\ o - C * - - *- - - C , I- 00 O
(A m U 00 o
I- C I O tl e O C
.0 S C . . . .. . . .. 1-
00~~~~~~~0
C)~~~~~~~~~~~
n ~ ~ ~ ~ LS oo -~~~~~~~.4o 00
U U o
22 +
; ~ ~~~ 00 X 0) Z 4. 00
i~~~~~- +- ;+I oO r C
) C S X S hS C C C C C 00 V II I
esz2C)C0)E2,,O4;422 ~
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
cwc~DN~ cD.-.-.; ,-coo .5 rAcC C
-o Q-* * o o 1- o0 s
Ln C) 00 10 I I I I 0o
L(q N0q N-N(Q DQ~ (N e((N .-0cI . . . -0~ :
((.0 11
tp.
C) . . 0~~~~~~~~~~~~~C b0<
oo eNN
4 "o Lr~ C-; s; i i oo oS
o~~***-//-/2Q
0 e ? . . . .0CA- - - -. - 00
b00
b-005(
Oq O D O0 oo O O00
00 .n rn ON 00 . . . . .0 cr, - L "C . .
-u u
0 00
~~~.0~t S- .0
O 0.o oo ? t O N Y a P4
2+-Ln A 0 io
0~t.
? m ^N tr~~~~() (i.'-( t)^E*~ _c
m u
.5oGi
?d~~~-(-)*})Ow>
?ci *?
23N * * m m m m tl oo e tl O m v) i,, Y
00 O O A u u H. . . . . . . . . . . . . . . O O O D~~~~~~~~~~~~~~c
7~~~~~~~
u~~~~~~~~~~~C
u~~~~~r u Ol C
>"4- ~ ~ ~ XX
u u 0?
uo u u} u O u tr4 ur O O a= O
N d O'i m ' N N / r f5
00e O00. .. 0 Om O? 23C
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
24 Cameron/Heckman
problems that arise from excluding workers on the basis of their labor
force or educational activity, we estimate a bivariate-selection-correction
model discussed at length in Cameron and Heckman (1992). Details of
the estimation procedure are available from us on request.7 The variables
used in the analysis are defined in Appendix A, although the common
English meanings are precise enough.
For all specifications of the wage and labor supply equations with and
without selection corrections, we are unable to reject the hypothesis that
GED recipients are indistinguishable from high school dropouts (see the
p values for the test of the hypothesis "GED = 0" given at the base of
table 9, pts. A and B). For all specifications of the labor-supply equations
and for specifications of the wage functions that exclude job tenure and
work experience, we reject the hypothesis that the GED degree is equivalent
to the high school diploma ("GED = HS GRAD"). When job tenure and
work experience are entered as regressors in wage equations, there is less
evidence of a distinction between the two forms of high school certification.
There is a strong negative relationship between total work experience and
GED status. In Cameron and Heckman (1992), we also document that
GED recipients are like high school dropouts and unlike high school graduates
in their job tenure and in their high levels of annual unemployment.
Using conventional statistical significance levels, the NLSY data strongly
reject the hypothesis that GED recipients are the labor market equals of
high school graduates. The same data do not reject the hypothesis that
high school dropouts and GED recipients are indistinguishable. A closer
look at the evidence indicates, however, that GED recipients lie between
dropouts and graduates in their economic standing but are much closer to
dropouts.
It is plausible that differences in economic outcomes among GED recipients,
dropouts, and high school graduates are largely due to differences
in ability. (Recall the ordering reported in table 2.) Table 10 presents
estimates of augmented versions of the models presented in table 9, part
A, when an AFQT test score-interpreted as a measure of ability-is added
to wage and hours of work equations. The test scores may be as much a
consequence as a cause of schooling, so the results shown in these tables
should be interpreted with caution. Introduction of the AFQT variable
tends to reduce the precision and size of the estimated GED and high
7 The selection-correction procedure used here does not play a central role in
producing these inferences. However, it does affect the strength of the inference
in the specifications of the wage function that include tenure and experience. In
Cameron and Heckmnan (1992), we examine the fit of estimated selection-corrected
and uncorrected wage and labor supply functions to the data. The selection-corrected
wage models fit the data, although the uncorrected wage models do not. Neither
corrected nor uncorrected hours models fit the data.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 25
school graduation coeffic
schooling. However, the F-tests reveal that the central inference of table
9, part A, is not reversed. The same is true for 28-year-olds (see table 10,
pt. B). GED recipients are statistically indistinguishable from high school
dropouts in terms of their hourly wages and hours of work and have lower
wages and hours of work than traditional high school graduates.
The evidence presented in tables 9 and 10 adjusts for differences in work
experience among high school dropouts, GED recipients, and high school
graduates. It implicitly assumes that a year of work experience has the
same effect on wages irrespective of educational attainment. If advocates
of the GED testing program are correct, GED recipients enter a new career
track after attaining their certificate. Such work experience is likely to have
greater training and wage-enhancing content than work experience obtained
as a high school dropout. Accordingly, the evidence presented in
tables 9 and 10 understates the contribution of the GED to lifetime earnings
and occupational advance by failing to recognize that GED recipients have
relatively fewer work experience years at the high school graduate level
than do traditional high school graduates.
Table 11 sheds light on this issue. It presents estimates of a wage equation
that. segments work experience by the years of educational attainment at
the time the experience was generated. The results indicate a high value
of work experience for high school dropouts who do not attain a GED,
but a low value of work experience for those who do. Post-GED work
experience produces virtually the same economic return as work experience
for dropouts who never attain the GED. Post-high-school-graduation work
experience produces a higher economic return, but it is not statistically
significantly different from the effect of dropout/post-GED work experience
on wages. Adjustment for work experience by educational attainment
does not reverse our conclusions.
Table 11 also indicates that, controlling for work experience, there is
little difference in the economic returns to GED certification or high school
graduation. These results reinforce a conclusion already gleaned from tables
9 and 10: that a major difference between GED recipients and high school
graduates is in labor supply and work experience, not in wages paid standardizing
for those characteristics. Consistent with the Milwaukee surveys
cited above (Pawasarat and Quinn 1986), employers are reluctant to hire
GED recipients.
Table 12 presents additional evidence on this point for the bulk of GED
recipients (those who receive their degrees between the ages of 17 and 19).
GED recipients are much less likely than high school graduates to be
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
0) e C
o-Q Q~- 0 0- - * * ~ 0
( O 00aQ-O Co * * * Q~ CQQ O oo
o o C2 - C C l o
5< Y -\ - U * * * * . . . . . . . .
~~~~~~~~~~C) Xn Ln C mmoot~~~~~~~~- ;g m I'l rl -- I
Mo U &n .N o 00 oo
5.J ~ 4 00 *t a * * ON O 0 oo
_~~~~~~~cnLq6OsCtCmeeOm n
CJ~ ~ i4O k n o
00.
s C 00 ( ,00 e000/m0000 m
0 t C o l O
tkE~~~~;ONnOn.O.O.O.O.O.O0.O.o<
o
(F 0
n) ;4 oui //
=~~~~~~~~~~~~~~~~~. b C0 tz
b-0
(n _ . _ aj t~~~~~~~~~~U m (O....D _r:?O?
.2 0~~~~~~
O o 0) ?0 .,.~m
o E; cK'oo omo ot?
Ct D X St 9 N--0(~t(0 - ,Xt . ' c
26
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
Lr C) :- :-4 . . . C - - - C) C)C ; )C
. . . . . . . . . .~~~~~~~~~~.
* i(~~~~~CI 0 L C
r) IC II I I
S
,,- C) C) CIA (D C IA --
r CN . . . . . n r 00 . 00
re I *I***I*I* O
Ci CD O N C N *O.
O 0 CD OCD 0 OON N O ? O C C X
Ln , ,C :I n CD Clrl C O0 GO n CD ON C 4X
0'tO-
* ~ 0 CA :- N I
CA Lr; r,~ r, O a
G: CD -Or r A On CDO t N 0 00 t5 ?
r.00 .- ( - --A C CA e * * * 00 n C- 0
I GO *l~* * O OSt OtNv
0 0
CD~~~~~~~~
- 00
0 0
0 ~~~~~~~~~~
0 0
-~~~~~~ 0~~~~
A U
C CIA 00 . . . . . . r. 00 :I OCOD? ._
U UE
7~~~~~2
t~~~~~~~~~~~~~~~~~o~~~~~
_E _E _E Y7 O O
:50 0.
C4~~~~~~~~~ 7
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
28 Cameron/Heckman
Table 11
Ordinary Least Squares Log-hourly Wage Regressions at Ages 25 and 28
(Year Effects Not Reported), Interacting Work Experience with Dropout,
GED, and High School Graduation Status and Breaking out Work
Experience for GED Recipients into Pre- and Post-GED Components
Age 25, Age 25, Age 28, Age 28,
No Selection with Selection No Selection with Selection
Intercept .437(6.8) .331(5.0) .655(5.8) .541(4.5)
GED .137(1.2) .110(-9) .262(1.4) .260(1.1)
High school graduate .141(2.1) .125(1.1) .334(2.8) .310(2.2)
Selection yC * ... .342(2.6) . .. .531(2.2)
Selection 72t . . . -.02(.1) ... -.190(-7)
Dropout experience .071(6.7) .072(6.3) .047(3.2) .047(2.6)
Post-GED experience .067(3.6) .067(3.2) .040(1.6) .038(1.2)
Pre-GED e erience -.035(1.3) -.028(1.0) -.045(1.3) -.045(.9)
Post high schooc
experience:: .075(9.8) .073(9.7) .031(3.1) .030(2.7)
Unemployment rate -.019(6.1) -.019(3.7) -.024(5.2) -.021(3.0)
2 years of college + GED .190(.9) .131(.4) .271(.2) .244(.1)
2 years of college + high
school .240(3.2) .210(2.3) .481(4.0) .420(3.3)
College graduate .354(5.3) .310(3.7) .522(4.4) .461(3.4)
Black -.128(5.8) -.100(2.5) -.142(4.2) -.091(1.9)
Hispanic -.020(.8) -.002(.0) -.000(-0) .017(.2)
R2 .17 .18 .16 .17
F-test: probability
GED =0 .21 .25 .15 .28
GED = high school
graduate .98 .98 .58 .86
Post-GED experience
= dropout experience .76 .86 .78 .89
Post high school
experience = dropout
experience = GED
experience .77 .88 .32 .54
NOTE.-Compare results for age
statistics are in parentheses and ar
for heteroscedasticity and estim
* Corresponds to the coefficient
in App. eq. (B2).
t Corresponds to the coefficient on selection correction term controlling for college enrollment in App.
eq. (B2) (available on request).
t A variable measuring experience before high school graduation was dropped; it was insignificant and
small in all specifications.
employed or in the military over the ages 20-28. Their labor force activity
resembles that of high school dropouts. These inferences are sustained in
the statistical tests reported in the final two columns of the table. Similar
results are found for those who receive their GED in their early twenties
(Cameron and Heckman 1992). GED recipients are not working the same
hours or acquiring the same work experience as high school graduates.
The observed ordering in economic status among dropouts, GED recipients,
and high school graduates may simply be due to differences in
years of schooling completed. Table 13, part A, reveals that on average
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 29
Table 12
NLSY Males: Proportion of Time during the Last Calendar Year Spent
neither Working nor in the Military for Dropouts and High School
Graduates and GED Recipients Who Received Their Degrees
between Ages 17 and 19 and Never Attended College
Wilcoxon Test: Wilcoxon Test:
High School GED High School > GED GED > Dropout
Age Graduate Recipient Dropout (p-Value) (p-Value)
20 .25(.01) .39(.02) .43(.01) .00 .16
21 .20(.01) .34(.02) .37(.01) .00 .21
22 .18(.01) .30(.02) .35(.01) .00 .06
23 .17(.01) .31(.02) .35(.01) .00 .05
24 .15(.01) .27(.02) .31(.01) .00 .21
25 .15(.01) .22(.02) .27(.01) .00 .14
26 .14(.01) .26(.03) .29(.02) .00 .38
27 .13(.01) .23(.04) .24(.02) .00 .36
28 .12(.01) .28(.06) .21(.02) .00 .69
NOTE.-Standard errors of the mieans ar
dropouts have completed one
Table 13, part B, establishes
completed 11 years of school
About 45% of the dropouts have 9 or less years of schooling compared to
only 10% of the GED recipients. If the ordering in labor market outcomes
among graduates, GED recipients and dropouts is simply due to years of
schooling completed, the value of high school exam certification as an end
in itself is in doubt. Government programs with such an emphasis are
misguided.
Table 14 sheds valuable new light on this question. That table displays
the effect on wages of interacting dropout and GED indicator variables
Table 13
School Completion for GED Recipients and Dropouts at Age 25
A. Mean Years of Secondary School Completed
Standard Error
N Mean of Mean
Dropouts 238 9.46 .08
GED recipients 125 10.40 .07
B. % Distribution of Years of School Completed
Years Completed
7 or Less 8 9 10 11
Dropouts 8.1 13.2 24.3 21.8 32.6
GED recipients 1.8 3.6 5.2 31.2 58.2
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
30 Cameron/Heckman
Table 14
Log-Wage Regressions Interacting GED and Dropout with Years
of Secondary School Completed (Year Effects Not Reported)
Age 25 Age 28
Intercept .613(17.6) .800(19.4)
Dropout after 10 years .100(2.1) .046(.9)
Dropout after 11 years .179(3.7) .120(1.5)
GED after 9 years or less -.156(1.7) -.020(.6)
GED after 10 years .113(1.6) -.010(.3)
GED after 11 years .191(3.9) .151(2.1)
High school graduate .225(7.5) .211(4.6)
2 years of college + GED .249(1.3) .152(1.0)
2 years of college + high school .318(6.3) .390(5.9)
4 years of college .421(11.2) .422(7.3)
Black -.189(8.9) -.160(4.8)
Hispanic -.040(1.6) -.024(.6)
R 2 .11 .12
F-test: probability > F:
1. GED 11 = dropout 11 .58 .49
2. GED 10 = dropout 10 .69 .76
3. GED 9 = dropout 9 .09 .52
4. High school graduate = GED 11 .62 .40
5. High school graduate = GED 10 .05 .04
6. High school graduate = GED 9 .00 .00
7. High school graduate = dropout 11 .24 .13
S. High school graduate = dropout 10 .01 .01
9. High school graduate = dropout 9 .00 .00
10. 2-year college and high school = 2year
college and GED .20 .13
11. Joint test: GED = dropout .52 .84
12. Joint test: GED = high school graduate
(excludes 2-year college) .00 .14
13. Joint test: GED = high school graduate
(includes 2-year college) .00 .12
NOTE.-Education dummn-iies are defined exc
t-statistics are in parentheses and use the M
with actual years of schooling
with 9 or fewer years of schooling. Dropouts with an additional year of
completed schooling earn 8%-1/0% higher wage rates. The same is true for
GED recipients holding postsecondary schooling constant. Differences
between GED recipients and dropouts are almost completely accounted
for by years of schooling. At the same completed schooling level, a GED
recipient earns only 1% more than a high school dropout. GEDs with 11
years of completed schooling earn only 3% less than high school graduates.
Dropouts with 11 years of schooling earn only 4% less than high school
graduates. Using the p values shown at the bottom rows of table 14, we
do not reject the hypothesis that GED recipients and dropouts with the
same years of schooling earn the same wages (see the first three rows of
the lower part of table). High school graduates and GED recipients or
high school dropouts with 11 years of schooling are also indistinguishable
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 31
(see rows 4 and 7). High school graduates earn statistically significantly
higher wages only compared to GED recipients or dropouts with 10 or
fewer years of schooling. Note further that high school graduates who
complete 2-year colleges earn 6% more than GED-certified males with 2
years of college, but this difference is not statistically strong (as measured
by p values). Too few GED-certified persons completed 4 years of college
to make a meaningful comparison at that education level.
Table 15 pushes the analysis of table 14 a bit further. When the total
number of years of schooling completed are added to the models in table
6, one cannot reject the Mincer (1974) specification that the coefficients
on the dummy variables indicating GED, high school graduation, and
various years of college certification are jointly insignificant at conventional
significance levels. There are no statistically precise "sheepskin" or "certification"
effects in the data controlling for the total number of years of
schooling completed. There is no cheap way to acquire the skills obtained
from conventional classroom instruction.
Cameron and Heckman (1992) present a parallel analysis for hours of
work. Again, years of schooling completed, not certification levels, account
for differences in labor-supply behavior.
The effects of adjusting for race, work experience, years of schooling
completed, local unemployment rates, and year effects on the wage distributions
of holders of different credentials at age 25 is depicted in figures
Table 15
Log-Wage Regressions at Ages 25 and 28, Controlling for the Total
Number of Years of School and College Combined
(Year Effects Not Reported)
Age 25 Age 28
Intercept .152(1.6) .509(3.4)
GED -.016(.7) .015(.4)
High school -.009(.3) .080(1.4)
2 years of college + GED -.094(.6) -.045(.2)
2 years of college + high school -.024(.4) .200(2.2)
4 years of college -.029(.4) .149(1.3)
Black -.188(8.9) -.160(5.0)
Hispanic -.034(1.3) -.015(.4)
Years of school .057(5.4) .034(2.4)
R 2 .11 .12
F-test: proba
GED = 0 .46 .72
GED = high school graduate .52 .21
Joint test: all education dummy variables = 0 .54 .16
Joint test: college education dummy variables = 0 .78 .16
NOTE.-Education dunmmies are defined exclusively. We failed to reject the hyp
that years of school have statistically different effects for years of college and
at ages both 25 and 28. t-statistics are in parentheses and are calculated using the M
to correct for heteroscedasticity.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
32 Cam eron/Heckrnan
6a-6c.8 Figure 6a shows the unadjusted figures for persons with no college.
Adjustments make both dropouts and GED recipients more like high school
graduates (fig. 6b). However, the mass of GED recipients is still concentrated
in the lower deciles of the high school wage distribution. The GED
recipient wage distribution is almost identical to the dropout distribution
except for the top decile (fig. 6c). The results for 28-year-olds show the
same pattern. GED recipients as a whole are much more like high school
dropouts than high school graduates.
D. Indirect Effects of Certification
The GED effects just discussed are partial or direct measures that hold
constant any effects of GED acquisition on postsecondary schooling and
training. The total effect of GED acquisition on wages also includes the
effect of certification on the volume of postsecondary schooling and training
multiplied by the return to this activity. Tables 7 and 8 reveal that GED
recipients are more likely to take postsecondary training and schooling
than are high school dropouts although they are less likely to attend and
complete such programs than are high school graduates.
Table 16 presents evidence on the indirect effect of GED certification
and high school graduation on wage rates. The wage equations reported
in table 9 are augmented to partition years of college completed more
finely and to include off-the-job training, apprenticeship and company
training, and military training as additional postsecondary training and
schooling choices.
Table 17 reports the components needed to estimate the indirect effects
reported in table 16 for 25-year-olds (results for 28-year-olds are available
in App. B, available on request). In the columns labeled "Estimated Returns,
the estimated effect of an extra unit of postsecondary schooling or
training on log wages is reported for GED recipients and high school
graduates. The rates of return to postsecondary activity for the two forms
of certification are statistically indistinguishable (see the first test at the
base of the table). However, by age 28, the returns to college for high
school graduates are higher than they are for GED recipients. With the
exception of military training, GED recipients take less postsecondary
training or schooling than high school graduates. The product of the rate
of return and the volume of training taken is the contribution of the form
of the postsecondary activity reported in each row to wages. The sum
across rows is the estimated total indirect effect. The estimated direct effect
is the coefficient on GED or high school graduation holding constant year
8 We use common work experience variables for all levels of education. This
appears to be justified in light of the evidence in table 11. Comparable figures for
28-year-olds are available on request in App. B.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 33
Frequency
251
20 -
15-
10 1
10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100
| - Drop Out - - GED Recipient FIG.
6a.-NLSY males, age 25, who have completed no college: how the GED recipient
and dropout hourly wage distributions fit into the high school graduate wage distribution,
by %. Source: NLSY (1979-87 data).
effects, postsecondary schooling, and dummy variables for race. The omitted
educational category is high school dropouts.
The indirect effect of high school graduation ranges between 34% and
42% of the total effect on wages. For GED recipiency, the indirect effect
ranges between 100% (at age 25) and 63% (at age 28) of the estimated
total effect. Although the estimated parameters for GED recipients are not
precisely determined, the evidence assembled in table 16 indicates the effect
of the GED on wages comes primarily through its effect on certification
for postsecondary training. The indirect effects for high school graduates
and GED recipients are nearly identical at age 25 and statistically indistinguishable
at age 28.
Frequency
25-
20-
15
lo
10
10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100
- Drop Out - l GED Recipient -l
FIG. 6b.-NLSY males, age 25, who have completed no college: how the GED recipient
and dropout hourly wage distributions fit into the high school graduate wage distribution,
adjusted for experience, race, unemployment rate, year effects, and the highest grade completed
Source: NLSY ( 1979-87 data). See table 9 notes for a definition of the sample.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
34 Cameron/Heckman
Frequency
251
20-
15 -
10 -
5
10 20 30 40 50 60 70 80 90 100
GED Recipient
FIG. 6c.-NLSY males, age 25: how the GED recipient hourly wage distribution fits into
the dropout wage distribution, adjusted for experience, race, unemployment rate, year effects,
and the highest grade completed, by 0/ Source: NLSY (1979-87 data). See table 9 notes for
a definition of the sample.
T he evidence reported in tables 6, 9, and 17 also weakly indicates that
the return to postsecondary schooling and training differs between high
school graduates and GED recipients. These differences are only partly
accounted for by the lesser amount of time spent in postsecondary education
by GED recipients. One possible source of these differences is the choice
of curriculum within each type of postsecondary education, but we have
no direct evidence on this possible explanation.
E. Some Longitudinal Evidence
Using the longitudinal structure of the NLSY, we compare a variety of
characteristics of GED recipients in the year before and after they receive
their certificate. Table 18 excludes persons in the military in the year before
or after receiving the GED.9 There is little evidence of any GED-induced
change in labor market outcomes in these tables, although the small sample
sizes may preclude precise determination of these changes.
Another way to study the impact of the GED on life-cycle wage growth
is to follow a cohort of young GED recipients over a period of time and
compare their experience to a group of young high school dropouts who
have not received the GED by age 25. Figures 7a-7f present the position
The exclusion of military personnel is done to avoid making pay comparisons
between military and civilian wage scales. However, inclusion of military personnel
does not affect our conclusions.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 35
Table 16
Direct and Indirect Effects of GED and High School Graduation
on Log Wages
A. Age 25 B. Age 28
High School High School
Graduate GED Graduate GED
College (combined) .0430(4.8) .0090(-8) .0890(5.4) .0118(.5)
Weeks of off-the-job training .0080(.3) .0150(.9) .0028(.7) .0190(-8)
Apprenticeship/company training .0130(5.5) .020(2.5) .0154(3.6) .0066(.3)
Military training .0040(1.2) .028(1.9) .0002(.3) .0036(.8)
Total indirect effect .068(5.4) .072(1.6) .017(4.6) .041(.6)
Total direct effect .129(5.3) -.003(-2) .142(3.6) .024(.5)
Total effect .197(5.2) .069(1.1) .249(4.2) .065(.6)
NOTE.-t-statistics are in parentheses. These numbers are
presented in the first column of table 9, pt. A-controllin
graduation, 2- and 4-year college education, year effects
year dummies-augmented with variables to control for
training, and completion of I year of college. Weeks of off
from a vocational-technical school, nursing school, flight
college. Weeks of military training are the total weeks of
duty, including active duty in the reserves. Weeks of offthe
college education indicators are interacted with dummi
to estimate separate returns for college, off-the-job-traini
and high school graduates. To construct the indirect effe
the estimated coefficient associated with each variable is
of off-the-job weeks and military training. To calculate th
for GED recipients and high school graduates for I year
college are multiplied by the return to each level of educat
The sum is the indirect effect of college education. The e
The total indirect effect is simply the sum of these com
estimated coefficient associated with the indicator for
total effect is the sum of the direct and indirect effects.
of GED recipients in the wage distributi
at ages 20, 22, and 25. The same persons are followed over time. The
unadjusted distributions (figs. 7a-7c) display the "Cosby effect," but by
age 25 there is great similarity in the distributions below the top decile.
The adjusted distributions are even more striking. By age 25, there is much
similarity in the two distributions below the median wage. Any initial
advantage of GED recipients below the top decile seems to have been
dissipated.
In Cameron and Heckman (1992), we document that GED recipients
are more likely to change jobs than are high school dropouts. Since a
significant portion of the wage growth of young men comes from job
changing, it is interesting to compare the wage growth of GED recipients
who change jobs after they receive the GED with the wage growth of
GED recipients who stay put. Results in Appendix B, available on request,
reveal that post-GED job changers receive some increase in wages while
job stayers receive little increase in wages. It is unclear, however, how
much of this growth in job changers' wages to attribute to job changing
and how much to attribute to receipt of the GED.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
36 Cameron/Heckman
Table 17
Direct and Indirect Effects of GED Recipiency and High School
Graduation on Log-Wages, at Age 25: Estimated Returns
and Sample Means
High School Graduates GED Recipients
Estimated Sample Estimated Sample
Returns* Meanst Productt Returns* Meanst Productt:
1 year of college .047(1.4) .10(.007) .005(1.3) .045(.5) .07(.019) .003(.5)
2 years of college .086(1.6) .11(.008) .009(1.5) .110(.9) .05(.016) .006(.9)
4 years of college .190(6.6) .15(.009) .029(6.3) N.A.? N.A. N.A.
Weeks of off-thejob
training .001(.4) 8.0(.7) .008(.3) .002(1.1) 7.6(1.6) .015(.9)
Weeks of
apprenticeship
or company
training .003(5.9) 4.0(.5) .013(5.5) .007(2.8) 2.9(1.0) .020(2.5)
Weeks of military
training .002(1.3) 2.2(.3) .004(1.2) .005(2.5) 5.5(1.3) .028(1.9)
Total indirect effect ... ... .068(5.4) ... ... .072(1.6)
Total direct effect ... ... .129(5.3) ... ... -.003(.2)
Total effect ... ... .197(5.2) ... ... .069(1.1)
Probability > F
Joint test: estimated returns for high school graduates
estimated returns for GED recipients .16
Joint test: sample means for high sc h0ol graduates
= sample means for GED recipients .00
NOTE.-Sumis and products may not appear exact du
* t-statistics are in parentheses and are constructed
t Standard errors of the mean are in parentheses.
t The variance of the product is calculated us
+ ,2var(iA) + var(O)var(jA), where var(O) is the varian
of the sample mean. There is no covariance since fB and
of the sample mean gives us the samle t-statistic for
this term as we do makes little if any difference in th
? There were no GED recipients who had completed college by age 25.
III. Reasons for Growth in GED Certification
The evidence presented in previous sections of this article suggest
the direct economic payoff to GED recipiency is low. If so, one mu
for explanations other than market benefits to account for the rapid
in GED recipiency.
The post-1963 growth in the proportion of high-school-certified persons
taking the GED evident in figure 1 is directly linked to the large-scale and
unprecedent expansion of the federal government and state programs in
human resources that began in the Kennedy-Johnson era. The two main
social programs that fueled the post-1963 growth in GED recipiency are
(a) the 1966 Adult Basic Education Act and subsequent amendments to it
and (b) a variety of federal programs for postsecondary education that
created a demand for high school credentials to qualify for program benefits.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 37
Table 18
Means for Those Working before and after Obtaining the GED
(for Individuals out of School and Holding a Civilian Job in the Year
before and the Year after Receiving the GED; N= 107)
Before After
Standard Error Standard Error
Variables Mean of the Mean Mean of the Mean
Hourly wage (1988 dollars) 6.18 .28 6.36 .28
Annual earnings (1988 dollars) 10,249.0 736.6 10,406.6 749.7
Annual hours 1,541.4 76.61 1,563.3 73.3
Annual weeks worked 38.0 1.46 37.7 1.44
Surprisingly, manpower-training programs that expanded greatly in the
1960s and 1970s contribute little to the growth in GED recipiency.
The Adult Basic Education Act of 1966 was a War on Poverty program
designed to provide adults with levels of education that were thought
likely to elevate them out of poverty. Throughout the course of the Adult
Basic Education program, the emphasis has shifted from an amorphous
goal of improving basic skills to a more easily specified and monitored
goal of producing GED-certified high school equivalents (DeSantis 1979).
Enrollment in this activity expands by a factor of five throughout the
period 1966-82.
Total expenditure on this program ceased to expand after 1973, and
the federal share in total program expenditure declines after that date.
Frequency
301
20-
10
0 -
10 20 30 40 50 60 70 80 90 100
FIG. 7a.-NLSY males, age 20
the dropout wage distribut
observations with nonmissing wages at ages 20, 22, and 25. Source: NLSY (1979-87 data).
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
38 Cameron/Heckman
Frequency
30 _
20 -
10 l
0 jm
10 20 30 40 50 60 70 80 90 100
FIG. 7b.-NLSY males, age 22:
the dropout wage distributio
87 data).
Figure 8 demonstrates that, in 1972, 24% of all GED recipients were produced
by Adult Basic Education programs, and the time series of GED
recipiency closely tracks the time series of GED credentials produced by
these programs. (These data are not available before 1972.) Amendments
to the 1966 act set forth in 1970 drop the age of eligibility for participation
in this program from 18 to 16 and add an explicit emphasis on high school
completion via the GED or by night school as a main objective of the
Frequency
30
20
10
0
10 20 30 40 50 60 70 80 90 100
FIG. 7c.-NLSY males, age 25: ho
the dropout wage distribution,
observations with nonmissing w
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 39
Frequency
30-
20 -
10
0
10 20 30 40 50 60 70 80 90 100
FIG. 7d.-NLSY males, age 2
the dropout wage distribution
and highest grade complet
observations with nonimissi
program. The amendm
the- reduced age requir
taking the GED. Most
school at least 6 months
periods for retaking th
instead of the previous
produced by Adult Bas
Frequency
30
20-
10
10 20 30 40 50 60 70 80 90 100
FIG. 7e.-NLSY males, age 22: how the GED recipient hourly wage distribution fits into
the dropout wage distribution adjusted for experience, unemployment rate, race, year effects,
and highest grade completed, by % (GED received by ages 17-19). Sample includes only
observations with nonmissing wages at ages 20, 22, and 25. Source: NLSY (1979-87 data).
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
40 Cameron/Heckman
Frequency
301
20 -
10-
0
10 20 30 40 50 60 70 80 90 100
FIG. 7f.-NLSY males, age 25: how the GED recipient hourly wage distribution fits into
the dropout wage distribution adjusted for experience, unemployment rate, race, year effects,
and highest grade completed, by % (GED received by ages 17-19). Sample includes only
observations with nonmissing wages at ages 20, 22, and 25. Source: NLSY (1979-87 data).
of all GEDs were trained by this program. Total enrollment increased
fourfold between 1970 and 1980.
Manpower training programs were introduced and expanded during the
early 1960s, beginning with the Manpower Development and Training
Act (MDTA) of 1962. The set of programs created by the act did not
emphasize academic training. Job Corps was an exception and did produce
GED recipients. However this manpower program was never large. In
600
500
400
300
200 ,_- - _
100
0
72 74 76 78 80 82 84 86
Year
Total -?----Adult Basic Ed
FIG. 8.-Total number of GED credentials issued and the number produced by the adult
basic education program (in thousands). Source: Council on Adult Education (various years);
GED Testing Service (1990).
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 41
1975, the number of Job Corps GED recipients was less than 2% of the
total granted. The successor programs to MDTA maintained its disinterest
in high school certification as a major objective and were negligible contributors
to the level or rate of growth of GED recipiency (Levitan and
Gallo 1989).
In addition to the growth in programs that made attainment of the GED
a main objective, there was substantial expansion in programs that required
high school degrees or their equivalents to receive benefits. These programs
fueled the demand for high school certification. Figure 9 charts the growth
in expenditure on major postsecondary educational funding programs that
required high school certification for eligibility. There is a corresponding
increase in numbers. There was gradual growth in National Defense Student
Loans, work-study support programs, and the Supplementary Educational
Opportunity Grant program during the period 1963-75 when GED certification
was growing steadily. All of these programs required a high
school degree or its equivalent for eligibility. Not only did the scale of
these programs increase over the period 1963-75, but their benefits became
applicable to less academically oriented postsecondary institutions such as
not-for-profit proprietary training centers.
The most dramatic development in postsecondary educational finance
was the growth in the Pell grant program in the period 1973-81 (see figs.
1 Oa and 1 Ob). Starting in 1973, benefits for all components of this program
could be used to finance proprietary training. Family-income restrictions
were relaxed, and loans became more widely available to the middle class
in 1976. Pell grants to proprietary students continued to grow after 1978,
3000
Four Year
2500 Two Year/
Proprietary
2000 !
1500
0
72 74 76 78 80 82 84 86
Year
FIG. 9.-Federal expenditures on major postsecondary education programs: Pell grants,
work-study, and Supplemental Education Opportunity grants, by type of institution (in
millions of 1988 dollars). Sources: U.S. Department of Education (1981); National Center
for Education Statistics ( 1990).
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
42 Cameron/Heckman
4000 Pell
Grant
National Direct Student Loan
3000 - College Work Study
ao Supplemental Education Opportunity Grant
T 2000-
.E/
o0 --
58 62 66 70 74 78 82 86
Year
FIG. lOa.-Federal expenditures on major postsecondary education programs (in millions
of 1988 dollars). Source: see fig. 9.
while payments to 2- and 4-year college students stabilized after 1976.
Between 1977 and 1981, guaranteed disbursements rose sharply with the
passage of new student loans amendments that allowed students at all
nonprofit and proprietary postsecondary institutions access to government
grants and loans to high school graduates and GED degree holders and
that liberalized family income restrictions on loan eligibility (see App. B
for figures, available on request). Annual commitments and annual par-
3000 Pell
Grant
2500 - National Direct Student Loan
College Work Study
2000 - Supplemental Education Opportunity Grant
l 1500-
1000 - - - - -
500 -
0 -
58 62 66 70 74 78 82 86
Year
FIG. lOb.-Participants in major federal postsecondary education programs (in thousands).
Source: see fig. 9.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 43
ticipants in the Guaranteed Student Loan program grew more than threefold
in these 4 years. A sharp rise occurs in the number of GED degrees
issued relative to all high school credentials during this same period
(fig. 1).
In 1979 and 1980, new regulations became operative that allowed any
individual with the "ability to benefit," including high school dropouts, to
participate in any of these programs. A General Accounting Office study
of proprietary institutions in 1984 found dropouts to be more likely than
high school graduates and GED holders to drop out from their programs
and more likely to default on loans and on grant obligations (U.S. General
Accounting Office, 1984, p. 56). Because of the threat of federal sanctions
imposed on institutions with loan default rates exceeding 15% for 2 consecutive
years, lending agencies had an incentive to screen out dropouts
so that GED status was still a valuable attribute for participation in these
programs.
Temporal coincidence can never establish causation. However, the close
association between the growth in GED recipiency and the growth in
government programs that subsidize attainment of the GED or require
high school certification for eligibility is strongly suggestive of an important
role for government subsidy policies in accounting for the growth in GED
certification (see fig. 1). This evidence helps to reconcile the growth in
GED certification and the low economic return to obtaining a GED that
we have documented in this article.
IV. Summary and Conclusion
Over the past 25 years, there has been dramatic growth in the propor
of high school credentials achieved by means of exam certification rat
than by the traditional route of high school graduation. The growth in
exam certification helps to reconcile the recent decline in the proportion
of 17-year-old high school graduates and the constancy in the proportion
of 20-24-year-olds with high school certificates. Exam certification is the
principal vehicle through which black and Hispanic high school certification
rates have approached that of whites. This article explores the causes
and consequences of this phenomenon.
The main conclusion of this article is that exam-certified high school
equivalents are statistically indistinguishable in their labor market outcomes
from high school dropouts. Both dropouts and exam-certified equivalcnts
have comparably poor wages, earnings, hours of work, unemployment
experiences and job tenure. GED-certified persons are closer to hi gh sc- ;. .'
dropouts than traditional graduates in their measured ability and in tilr
market status. Even after controlling for ability, GED-certified males have
inferior labor market status compared to high school graduates. GEDs
have lower employment rates and less work experience than high school
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
44 Cameron/Heckrnan
graduates. Both anecdotal and empirical evidence suggests that employers
and the military discount the GED.
This conclusion is strengthened when account is taken of years of
schooling completed. Whatever difference is found among GED recipients,
dropouts and high school graduates is largely accounted for by years of
schooling. There is no cheap substitute for classroom instruction. Educational
programs that focus on the GED as an end in itself are misguided.
Whatever economic return exists from GED recipiency arises from its
value in opening postsecondary schooling and training opportunities. GED
recipients take less postsecondary training than high school graduates
(military training is an exception to this rule), and they receive lower
returns-especially for their college education. The available evidence indicates
that GED recipients who attend college take a more vocationally
oriented curriculum than high school graduates. Both anecdotal and
econometric evidence suggests little direct market value for the GED controlling
for returns from postsecondary training.
An important qualification to our analysis should be stated. The sampling
frame of the NLSY has forced us to confine our attention to the early
stages of adulthood. It is possible that GED recipients and high school
dropouts will look more dissimilar at older ages and that GED recipients
and high school graduates will look more similar. That issue can only be
settled by looking at later waves of the NLSY data or by using other data
sources with older persons. We are currently engaged in that task.
Since the economic value of GED recipiency is low, its recent dramatic
growth as a means of high school certification is apparently paradoxical.
Our investigation of the political economy of the GED resolves this paradox.
Federal and state Adult Basic Education programs subsidize GED
test taking and use GED recipiency as a measure of monitoring bureaucratic
performance in these programs. The growth in funding and participation
in these programs tracks the time series of GED recipiency closely. In
addition, over the past 25 years, there has been dramatic growth in the
federal subsidy to postsecondary schooling and training programs. High
school certification is a requirement for participation in these programs.
This subsidy has created a derived demand for GED certification.
The dramatic rise in GED certification is a consequence of federal and
state government policies. The direct subsidy to certification and the derived
demand for GED certification in order to receive subsidies for postsecondary
training reconcile the low gross economic returns to certification
and the rapid growth in GED recipiency.
Our study sheds new light on the value of psychometric test scores in
predicting labor market outcomes. Much of the debate about the success
and failure of public and private schools focuses on psychometric measures
of cognitive ability. Our evidence on the irrelevance of successful exam
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 45
performance on labor market success and success in postsecondary education
suggests that current evidence in the private schooling debate is of
little relevance for gauging the importance of schooling organization on
the long-term economic success of students.
Appendix A
The National Longitudinal Survey of Youth and Description of
Variables
This appendix contains a brief description of the NLSY and the
variables used in the analysis of wages and labor supply. Appendix B
contains a full description and is available from us on request.
The micro data we use are from the 1979-87 waves of the National
Longitudinal Survey of Youth. The NLSY includes a randomly chosen
sample of 6,111 U.S. youths and a supplemental sample of 5,296
randomly chosen black, Hispanic, nonblack, non-Hispanic, economically
disadvantaged youths. The youths were ages 14-21 in 1979 and were
interviewed annually beginning in 1979. Our sample consists of males
who were in the random sample, the black supplemental sample, and
the Hispanic supplemental sample. We have a total of 4,837 individuals.
To examine the effects of having a GED or high school diploma on
hourly wages and labor supply, we take a subset of our data, which
were sampled at ages 25 and 28. For 25-year-olds, we include everyone
between ages 16 and 20 in January 1978. Altogether, 3,139 individuals
from the random, Hispanic supplement, and black supplement are
interviewed at age 25. For our study of wages at age 28, we could
include only those ages 19 and 20 and a portion of those who were age
18 in January 1978, for a total of 1,284. Of these, approximately 6.5%
were dropped at each age because of missing values in the job tenure
variables or because hourly wages were greater than $60 or less than
$1.50 (1988 dollars). Our sample has 2,926 males age 25 and 1,199 males
age 28. If an individual was enrolled in college during the past survey
year, he was excluded from our analysis of wages. Those who were
counted as unemployed or out of the labor force for a reason other than
school attendance were those with no job during the survey year who
were not in school. These individuals were also excluded. Definitions of
all the variables used in this analysis should be straightforward. The
following is for added clarity:
Hourly wage Hourly wage is 1988 dollars at the current
or most recent Job
Annual earnings Earnings for last year in 1988 dollars
Annual weeks, annual Total weeks or total hours worked last
hours calendar year
Tenure Tenure in weeks at the current or most
recent job
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
46 Camreron/Heckm an
Experience Total experience in weeks excluding weeks
worked in high school. Weeks at the
current or most recent job since the
individual was 16 years old were
subtracted only when this variable was
used in regressions.
References
Beder, Hal. "What Has Happened to Iowa's GED Graduates?" Research
report. Des Moines, Iowa: Department of Education, April 1992.
Bickel, Peter, and Doksum, Kjell. Mathematical Statistics: Basic Ideas and
Selected Topics. San Francisco: Holden-Day, 1977.
Cameron, Stephen, and Heckman, James. "Determinants and Educational
Consequences of High School Dropout and GED Certification." Unpublished
manuscript. Chicago: University of Chicago, June 1991.
. "The GED." Unpublished manuscript. Chicago: University of
Chicago, Summer 1992 (revised from a June 1991 draft).
Chubb, John, and Moe, Terry. Politics, Markets and America's Schools.
Washington, D.C.: Brookings Institution, 1990.
Coleman, James; Hoffer, James; and Kilgore, Sally. High School Achievement.
New York: Basic, 1982.
Council on Adult Education. Participation in Adult Education. Washington,
D.C.: U.S. Department of Education, various years.
DeSantis, Vincent. "The Adult Education Act, 1964-1979: A Political History."
Upper Montclair, N.J.: National Adult Education Clearinghouse,
1979.
Efron, Bradley. The Jackknife, the Bootstrap and Other Resampling Plans.
Philadelphia: Society for Industrial and Applied Mathematics, 1982.
Finn, Chester E. "The High School Dropout Puzzle." Public Interest 91
(Spring 1987): 173-92.
GED Testing Service. The 1989 Statistical Report. Washington, D.C.:
American Council on Education, 1990.
Katz, Lawrence F., and Revenga, Ana L. "Changes in the Structure of
Wages: The United States vs. Japan." Journal of the Japanese and International
Economies 3 (December 1989): 522-53.
Kominski, Robert. "Estimating the National High School Dropout Rate."
Demography 27, no. 2 (May 1990): 303-11.
Laurence, Janice. "The Diploma as a Military Performance Predictor: It
Works, but Why?" Human Resources Research Organization. Paper
presented at the ninety-first annual convention of the American Psychological
Association, Anaheim, Calif., August 1983.
Levitan, Sar A., and Gallo, Frank. A Second Chance. Kalamazoo, Mich.:
Upjohn Institute, 1989.
Lindley, David V. "A Statistical Paradox." Biometrika 44 (1957): 187-92.
McKinnon, James, and White, Halbert. "Some Heteroskedasticity-consistent
Covariance Matrix Estimators with Improved Finite Sample Properties."
Journal of Econometrics 29 (1985): 305-25.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms
GED Nonequivalence 47
Malizio, Andrew, and Whitney, Douglas. "Who Takes the GED Tests? A
National Survey of Spring 1980 Examinees." GED Testing Service, Research
Studies no. 1. Washington, D.C.: American Council on Education,
March 1981.
. "Examinee and High School Senior Performance on the GED
Tests: A Detailed Analysis." GED Testing Service, Research Studies no.
3. Washington, D.C.: American Council on Education, June 1982.
Mincer, Jacob. Schooling, Experience and Earnings. New York: Columbia
University Press for the National Bureau of Economic Research 1974.
National Center for Education Statistics. The Digest of Education Statistics.
Washington, D.C.: U.S. Department of Education, 1989.
Pawasarat, John, and Quinn, Lois. "Research on the GED Credential and
Its Use in Wisconsin." Wisconsin State Department of Public Instruction,
Employment and Training Institute, research report. Milwaukee: University
of Wisconsin, August 1986.
U.S. Bureau of the Census. Current Population Reports, Series P-20. Washington,
D.C.: Government Printing Ofice, various years.
U.S. General Accounting Office. Many Proprietary Schools Do Not Comply
with the Department of Education's Pell Grant Program Requirements.
Washington, D.C.: Government Printing Office, 1984.
This content downloaded from 147.251.185.127 on Mon, 05 Mar 2018 14:34:56 UTC
All use subject to http://about.jstor.org/terms