«
f
UBRAJBBS 1
§
0€WE
HD28
.M414
no-
mi
-
ALFRED
P.
WORKING PAPER
SLOAN SCHOOL OF MANAGEMENT
THE IMPACT OF PRIVATE SECTOR TRAINING ON RACE
AND GENDER WAGE DIFFERENTIALS AND THE CAREER PATTERNS
OF YOUNG WORKERS
by
Lisa M.
WP# 3353-91 -BPS
Lynch
November, 1991
MASSACHUSETTS
INSTITUTE OF TECHNOLOGY
50 MEMORIAL DRIVE
CAMBRIDGE, MASSACHUSETTS 02139
THE IMPACT OF PRIVATE SECTOR TRAINING ON RACE
AND GENDER WAGE DIFFERENTIALS AND THE CAREER PATTERNS
OF YOUNG WORKERS
by
Lisa M.
WP# 3353-91 -BPS
Lynch
November, 1991
UBRARIEt
DEC 2 1991
RECEIVED
M.t.T.
FINAL DRAFT REPORT
Department of Labor grant number E-9-J-9-0049
The Impact of Private Sector Training on Race and Gender
Wage
Differentials
and the Career Patterns of
Young Workers
by
Lisa M. Lynch*
M.LT. and NBER
Submitted
July, 1991
thank the participants of seminars at Boston University, Columbia, M.LT.,
Northwestern, LSE, NBER, the Sloan conference on the Longitudinal Analysis of Labor
Market Interventions at Yale University, and the Department of Labor for helpful
comments on a previous draft. I also benefitted from comments by Dan Black, James
Heckman, Peter Kuhn, Edward Lazear, Jacob Mincer, Michael Pergamit, Fabio
Schiantarelli, Nachum Sicherman, James Walker, and Andy Weiss. Thorough research
support was provided by Andrea Ichino and Pam Loprest. This project was funded by the
U.S. Department of Labor, Bureau of Labor Statistics imder Grant Number E-9-J-9-0049.
I
would
like to
Opinions stated in
this
document do not
necessarily represent the official position or policy
of the Department of Labor.
Note: Sections of
this report
have appeared in Lynch (1991a) and Lynch (1991b).
TABLE OF CONTENTS
P^® *
SECTION
^
Summaiy
L
Executive
n.
Introduction
in.
Private Sector Training and Wages: Theoretical
'
^"
Framework and Data
IV.
Private Sector Training and Wages:
Empirical Results
19
V.
Private Sector Training and Wages:
Conclusions
28
VI.
Private Sector Training and Mobility: Theoretical
^^
Framework and Data
Vn.
Private Sector Training and Mobility:
Empirical Results
38
Vm.
Private Sector Training and Mobility:
Conclusions
42
DC.
Footnotes
X.
References
XI.
Tables
45
48
51-65
I.
Executive
Summaiy
Objectives
While the returns to a college degree or government training programs
have been widely documented, there has been
in the U.S.
relatively little analysis of the returns to other
forms of human capital investment that non college graduates undertake. This has been due
However, using the
primarily to the lack of appropriate data for this type of analysis.
unique features associated with the National Longitudinal Survey Youth Cohort, NLSY,
how
study analyzes
personal characteristics including employment histories and local
demand conditions determine
the probability of receiving training and their effects
wage growth, and employment mobility of workers. More
addressed
here
this
include
the
relative
importance
specifically,
some of
and tenure
of training
on wages,
the issues
for
wage
determination and the rate of return to company provided training versus training from forprofit proprietary institutions
from employer
union
to
status, race
and regular schooling. The
employer and the existence of
portability of
company
training
differentials in the returns to training by
and gender are also investigated.
Finally, the
impact of different types
of training investment on the probability of leaving an employer are examined.
Methodologv
Using a standard human capital framework
rates of return to training, schooling
this report first
and tenure by race and gender for young workers. To
control for possible selection issues arising with the
training,
estimates the varying
non random
two stage estimation procedures are applied. In addition a
probability of receiving
first
difference equation
is
estimated to control for the effect of unobserved time invariant individual fixed effects.
The study then presents
estimates on the impact of different types of training on the
probability of leaving an employer using a
Cox proportional hazards model with time varying
covariates.
Findings
This report shows that private sector training plays a significant role in the
^
determination of wages and wage growth of the 70 percent of young workers in the U.S.
do not graduate from
college.
Specifically,
when
private sector training
different types (on-the-job training, off-the-job training,
different patterns emerge.
For example, the
much
less likely to receive training within
divided into
and apprenticeships) some very
characteristics that
probability of receiving training are primarily race
is
who
and gender.
appear to influence the
Women
and nonwhites are
a firm either through an apprenticeship or other
forms of on-the-job training. This differential pattern in the acquisition of training by race
and gender may be a
partial explanation of the persistent
wage gap between males and
females and whites and nonwhites. Schooling raises the probability of receiving off-the-job
training
and apprenticeships but
it
had a smaller impact on the probability of receiving firm
provided on-the-job training.
^ All types of training raise
wages
significantly. In particular, this
paper shows that on
average, for this sample of non-college graduates, off-the-job training from proprietary
institutions
(1989).
can be useful for increasing wages. This
The prime
is
in contrast to
difference between this study and that of Leigh
for off-the-job training to
a recent study by Leigh
is
that this study allows
have an effect not only on current wages but on future wages as
well.
In addition,
it
shown
is
While workers receiving
spells they are
to
more
that longer training spells have a larger effect
may
ofif-the-job training
on wages.
receive lower wages during their training
use that training to leave their low wage employer and
likely to
move
a better paying job.
Finally, while on-the-job training with the current
employer increases wages with the
current employer, this type of training seems to be quite firm specific since on-the-job
training
from a previous employer
is
some evidence
there seems to be
workers on the job
it is
for those
never significant for current wages. At the same time,
that
if
general training
who have
is
being given to any group of
not completed high school.
Implications
While
this project
has attempted to shed
young workers and the consequences of
are
still
many
issues that
the duration of the
first
on
this
new
light
their
on the
skill
formation process of
wages and patterns of mobility there
remain unresolved. This paper has modeled the determinants of
job after school, not subsequent employment.
future research should examine for example
differences change over time.
It
would
how some
also
As
the
NLSY
age
of the gender, race and educational
be interesting to examine the hazard rates by
broad industry and occupational categories.
robust the findings are after additional work
Finally,
is
done
it
would be important to see how
to address the endogeneity issue for
training.
Nevertheless, there
workers and private sector
for
young workers
is
a story that emerges from the results in this report for young
training.
in their first job.
Company
training in the U.S.
Young workers
is
very firm specific, even Q
entering the labor market can receive
both 'good' and 'bad' draws from the labor market. There are some workers who get a 'bad'
draw who appear
in 'good' jobs are
to
move
more
to better
findings
investing in off-the-job training.
likely to obtain on-the-job training
a lower probability of leaving the
The
employment by
firm.
These
finding that on-the-job training
from the Hudson
Institute
is
which
results in higher
wages and
effects are particularly strong for
primarily specific
which surveyed 645 firms
is
Those
women.
consistent with recent
in the U.S.
and found
that only
8 percent had any sort of general remedial on-the-job training programs^^ The fact that
U.S. firms are
more
willing to invest in firm specific training than in general training
is
imderstandable given the inability to "capture" the returns on investments in general training.
However, whether or not U.S. firms
will
be able
the future, given the characteristics of the
demands of new technology,
is
new
questionable.
to
remain competitive with
this strategy in
entrants into the workforce
and the
skill
Introduction
II.
While there have been numerous studies devoted to examining the impact of
governmental training programs on workers
market,
has been remarkably
there
who have
little
experienced
research
difficulties in the labor
on the actual occurrence and
consequences of training provided by the private sector. Since one possible explanation of
the lower productivity growth in the U.S. relative to countries such as
is
that firms in the U.S. underinvest in their workers,
imderstanding of the
human
capital
strategies
it
Germany and Japan
have a better
crucial to
is
of firms and workers
and of
their
consequences.
Obtaining an estimate on
how much
private sector in the U.S., however,
is
is
currently being spent
extremely
on
training by the
determine.
difficult to
It
has been
estimated by Camevale (1986) that $150 billion are spent annually on K-12 education, and
€'
as
much
sector.
as $210 billion are spent annually
Approximately $25
their first
billion of the
Training Magazine in
job\
more employees, reported
$210
its
,
on formal and informal training by the private
billion are spent
annual survey of training by firms with 100 or
that in 1988 over $45 billion
Bartel (1989) reported an even larger
on young workers entering
number of $55
were spent on formal
billion for
training while
formal training from the
private sector in 1987 using firm survey data. Finally, Mincer (1989) calculated that as
as $148 billion
may have been
individual data.
training for
much
spent on formal training programs by employers in 1987 using
Therefore, given the $25 billion or
yoimg workers, the
issue
is
more spent by
the private sector on
not that U.S. firms do not invest in their workers,
but rather that the nature and the size of these investments
may
not be enough for the
new
entrants in the 1990's.
The
difficulty in
documenting the actual investment in training
representative
large part to the lack of a comprehensive,
and
their
human
receives training,
management
resource
what types of
training
poUcies.
As a
in the U.S. is
result,
we know
Uttle
test
human
effect of training
profiles.
have
little
to
Apart from the
to infer the
impact of training on wages from
fact that this is
a rather unsatisfactory way to
The
more than 70 percem
of young workers
who do
Where do young workers who do
Is
it
Are they able
wage
through schooling
What happens
to those
8
(1989) as
development process
for the
new jobs and
realities of
for-profit proprietary business
young people who do not even
to obtain the necessary general
productive workers?
in
been documented
Bloom and Freeman
skill
-
not graduate from college acquire training
from on-the-job training or from
vocational institutions?
rising
not finish coUege. Yet, these are the
young workers who are viewed by many as being unready
after school?
is
returns to a coUege degree have
about the
examples of recent papers) but we know relatively UtUe
school?
which imply
training
of the primary ways young workers acquire
Blackburn,
by many, (see Katz and Revenga (1989) and
the 1990's.
on the wages
do with productivity enhancing uaining.
particular by completing coUege.
of the
who
on the probabiUty of remaining with
theories
capital theory, there are several alternative
profiles that
One
wage
about
programs are provided and where, the degree of
an employer. Consequently many have had
the shape of
in
and longitudinal survey of firms
the impact of training
firm specificity and portability of firm provided training,
and wage growth of workers, and the
due
and
and
finish high
specific skills training to
become
Several studies have attempted to use various measures of private sector training and
examine the in^)act of training on wages
shape of wage
profiles.
directly rather than inferring the effect
These studies include Mincer (1983, 1988), Brown (1983, 1989),
and Tan (1986), Pergamit and Shack-Marquez (1986), and Barron cL
Lillard
from the
However, each of these papers
is
subject to different limitations.
issues include the lack of complete
Of the more
employment, training and schooling
individuals in the various surveys, difficulties in measuring the
training the respondent received,
Some
and problems
al.
amount of
(1987).
critical
histories
on
private sector
in distinguishing firm-specific
from general
types of training.
It is
into
possible, however, to
overcome many of these problems and gain new
private sector training in the
Longitudinal Survey youth cohort,
one to reconstruct
the
moment
insight
U.S. using longitudinal data fi-om the National
NLSY.
This data
set,
despite
limitations,
its
for the first time the entire formal training history for
does allow
an individual from
they enter the labor market. This event history includes both the occurrence
and duration of each training
spell.
Given the current debate about the need
for
more
knowledgeable workers in the 1990's to deal with the demands of new technologies, firms
may be
increasingly required to switch
formal training programs.
An
from informal on-the-job training to more structured
empirical analysis of formal training programs and of their
consequences, therefore, would be useful. Moreover, the
to distinguish
between
training, training
While
it
from
is
NLSY
data allow the researcher
different sources of private sector training
for-profit proprietary institutions,
- company
provided
and apprenticeships.
not possible in a single report to investigate
all
aspects of training
mvestments for young workers,
employment
and
training
More
histories
their effects
specifically,
training
and
local
for
training versus training
demand
how
receiving
conditions determine the probabiUty of
relative importance of
issues addressed here include the
wage determination and the
from
rate of return to
for-profit proprietary institutions
portability of company training
in the returns to training
personal characteristics including
mobility of workers.
on wages, wage growth, and employment
some of the
and tenure
this study analyzes
company provided
and regular schooling. The
of differentials
from employer to employer and the existence
by union
status, race
and gender are also investigated.
Finally, the
on the probabiUty of leaving an employer
impact of different types of training investment
are examined.
Theoretical
Private Sector Training and Wages:
III.
Many
wage
theories have
slope
profiles
been advanced
upwards.
Framework and Data
to explain individual variation in
(1974)
According to Becker's (1964) and Mincer's
as human capital or
fundamental work, wage profiles slope upwards
experience.
is
on wages
and consequently
in the
in their earnings.
form of a premium paid
not portable, the size of the
training.
skills
inaease with
there should be an increase in
Therefore, as workers acquire more training
their productivity
effect
wages and why
Firm
specific training will
have some
training
to reduce turnover, but since specific
premium may not be
as large as that paid for general
part
of training on wages will depend in
Therefore, the magnitude of the impact
on the degree of specificity of the
For example,
would pay
in a standard
training received
human
capital
for general training while firms
and
in part
on who pays
model one would expect
that individual workers
would pay for and provide firm
10
for the training.
specific training.
Some
firms
may provide general
training but in this case
you would expect to observe wages
investment by paying
negatively related to training as finns finance this general training
At the completion of general
workers a lower wage.
wages
rise
minimiim wages, however, may cause firms to be unable
substantially. Legislated or social
to reduce
wages should
trai nin g,
sufficiently to cover these general training costs
and consequently make
firms reluctant to provide general training.
While human
less
capital theory provides
one explanation about why wages are more or
sloping wage
upwardly sloping for workers there are alternative explanations of upward
profiles that
discuss
how
workers.
have
litUe to
do with
training.
firms offer upward-sloping
Specifically, Stiglitz (1975)
wage
and Lazear (1981)
profiles to discourage "shirking"
among
An alternative explanation (see Salop and Salop (1976) and RothschUd and Stiglitz
(1976)) might be that firms use
employment elsewhere.
upward sloping
profiles to discourage "movers"
from seeking
Recent papers by Abraham and Farber (1987), Altonji and
matching
Shakotko (1987) and Topel (1987) have examined the importance of job
in
These studies have examined whether or not
(in
explaining
upward sloping wage
the absence of data
on
profiles.
training) the inclusion of tenure in a
wage equation simply measures
workers in long jobs are
job specific returns (such as training) or captures the fact that
worker-employer matches.
either better workers, in better jobs, or in better
measure of job-match quality
coefficient
on tenure
is
is
not included in the estimation then
it is
If
some
argued that the
biased upwards.
mutually
These alternative models of compensation should not be viewed as
exclusive; the
most
likely case is that
compensation
11
is
afferted by
some combination
of
human
capital
and other
The purpose of some of
factors.
the recent studies
on wages,
however, has been to show that after controlling for job match quality, the impact of tenure
or seniority on wages
is
small and to infer from this that
capital investments such as
But without detailed
have a negligible role in the determination of wages.
trainin g
information on the type of training undertaken,
human
human
capital investments
and whether they
it is
difficult to sort
out the real returns to
reflect general or specific capital, or other
factors.
Since the
spells across all
NLSY
data on training specify starting and ending dates of
employers
it is
spells of on-the-job training,
employer. In
human
if
employer provided training
this type of training
and imcompleted
wages since employers and employees
with specific training. Therefore,
if
will share
ON-JT
is
is
primarily general then
should be lower during a training spell and
Specific training, however, will have
an ambiguous
effect
positive.
However,
ON-JT
if
including a measure of
primarily general then the coefficient on an
The impact of
with a previous or current employer on wages should be
"better"
ON-JT
workers are more
in a
wage equation without
controlling for the selection of
upward bias on
training coefficients. This means, for example, that a significant
from a previous employer may be due
12
ON-JT, then simply
likely to receive
these "better" workers into training will result in an
training
on current
both the costs and the returns associated
interrupted spell of training with the current employer should be negative.
a completed spell of
training
ON-JT, from a current employer and ON-JT from a previous
capital theory
wages for workers receiving
higher afterwards.
possible to distinguish between completed
all
and
all
of the estimated
positive coefficient
to selection bias or evidence that
on
employer
provided training
insignificant
is
general and portable for young workers.
If instead this coefficient is
then the training is not portable and, therefore, suggests that
it is
primarily firm
specific'
Being able to imderstand the degree of
specificity
of firm provided training
particular importance in judging policies that subsidize employers
If
the subsidy
that
is
is
provided
is
be important
quite general, then
it is
important to see whether or not
in deciding the level of
Japanese employers invest substantially in general
finding out that private sector training in the U.S.
As noted above, before examining
A test of this assumption
is
training, especially for
European and f
younger workers,
quite firm specific reveals an interesting
its
competitors.
the impact of training
on the wages of young
necessary to examine the characteristics of those individuals
receive training. This
is
interesting in
its
an
this is in fact
In addition, given that
difference in the nature of training between the U.S. and
it is
young workers.
government support and the degree of monitoring
of employer provided training for young workers.
workers,
hire
of
motivated by a belief that when employers hire yoimg workers the training
appropriate characterization of firm provided training in the U.S.
will
who
is
own right and because
of sample selection bias in the wage equation.
The
it
who
actually
helps in tackling the issue
selection bias in the
wage equation
is
very similar to the "treatment selection" problem in the evaluation of the effectiveness of
government training programs (see Lalonde (1986) and Heckman and Hotz (1989)
excellent surveys
on
this).
If individuals
actual return to training in a
for
are not randomly assigned to training then the
wage equation may be biased upwards
controlled for.
13
if this
selection
is
not
In order to
that there are
training
~
individuals
job, total
model the
acquisition of private sector training
it is
important to realize
two possible agents who may influence the probability of a worker receiving
the individual worker and/or the firm. Finns are
who
they believe will be
more attached
more
to the firm.
likely to invest in those
Therefore, tenure on the
work experience, educational backgroimd and other demographic
are expected to influence the firm's investment decision. For example, firms
characteristics
may decide not
to invest in advanced training for their female employees because they believe that
are
more
likely to leave the firm.
If
they leave early in their tenure with the firm the firm
would not have sufGcient time to recoup
its
training investment.
In addition as Lazear
(1979) has discussed, the narrowing of the black/white and male/female
since the passage of affirmative action legislation
form of discrimination that resulted
of
wage growth. In other words,
by paying higher wages to
but wage growth would be
Individuals
who do
in
differentials
may have been accompanied by a different
a widening gap in the job-experience induced rate
blacks they
may have
at the
to these groups. Therefore, starting
much
wage
as employers responded to affirmative action legislation
women and
amount of training provided
women
same time reduced
the
wages might be the same
slower due to less training investment.
not receive training within the firm due to direct discrimination
or statistical discrimination may respond by obtaining ^dsible off-the-job" training to improve
their productivity
and opportunities
in other firms.
This individual investment in training
is
some
evidence of this type of behavior in the schooling decisions of blacks (see Lang and
Ruud
could also be used as a signal of their cormnitment to the workforce.
(1986)).
14
There
and industry will also affect firms'
Technological requirements of the occupation
change are more
size is also
likely to
expeaed
Larger firms
characterized by rapid technological
industries or occupations
THose
training decisions.
need to provide
skills training
receiving
to influence the probability of
may have better developed internal labor markets
and development of employees
and Tan
(see lillard
(1986)).
company provided
that rely
larger size
in the firm. In addition, the
on
NLSY
of training
so this important determinant
is
training.
internal training
may
Unfortunately information on firm size
marginal costs of training workers.
cveiy year in the
Firm
also lower the
is
not coUected
not included in
this
training.
In
analysis.
Finally,
schooUng may
affect
an individual's probabiUty of receiving
schooling
particular, additional years of
and aptitude
On the other hand, workers with poorer initial
in learning.
years of schooling
may
coUege graduates
it
interest
may signal a certain "stick-to-it-ness" and an
require additional training to get
will
be
up
skills
due to fewer
to speed. In this study of non-
the importance of finishing
particularly interesting to observe
acquiring employer provided training.
high school for the probabihty of
been Umited by
in wage determination have
Previous studies on the role of training
for analysis.
the nature of the data available
To
highlight
some of these problems Table
1
commonly used. Very
in a selection of surveys most
shows the differem questions contained
acquired on the
about the training the respondent has
few of these questions actually ask
the question used by
current or past jobs. For example,
training
of Income Dynamics, PSID, on
qualified for the job, not
how long
is
how
long
it
Brown
(1989) from the Panel Study
took the "average" person to become
the respondent actually took to
15
become
qualified. In the
older
MLS
cohorts analyzed by Mincer (1983, (1988)) and Lillard and
coUected relate to training received or used on the current job.
when
One
Tan
is
(1986), the data
not able to observe
the training actually took place or whether other types of training had been
undertaken by the respondent
received
is
Incomplete information on the total amoimt of training
also a limitation with the Current Population Siuvey,
CPS, data used by Pergamit
and Shack-Marquez (1986). The CPS questions are unlikely to provide information on the
training experience of older workers
training
if this
was acquired from previous employers.
Therefore, cross sectional analysis of the impact of training on wages using the
CPS data will
The Employment Opportimity
Pilot Project,
have to carefully control for cohort
EOPP,
data used by Barron et
al.
effects.
(1987, 1989) are interesting since they provide a
measure of the "representative" length (and
costs) of training to
an employer. However, the
data collected are restricted to information on the most recent hire in the firm.
recent hire
is
more
likely to
be
training investment observed
in the firm receive.
not representative of
is
in a position of high job turnover then
it is
If
the most
possible that the
an underestimate of what more "representative" employees
In addition, the
all
good
EOPP
firms in the U.S..
firms were predominately low
Many
wage
firms and
of these limitations are overcome with the
new NLSY.
The NLSY
at the
is
a survey of 12,686 males and females (who were 14 to 21 years of age
end of 1978) and contains detailed data on education,
programs marital
status, health
jobs, military service, training
and attitudes of young workers. The respondents have been
interviewed every year since 1979 on
all
aspects of their labor market experience.
response rate in 1985 was over 95 percent of the original cohort.
16
The data on
The
types of
training (other than governmental training or schooling) received are
some of
the most
comprehensive data available on private sector training. Respondents were asked about
what types of training they had received over the survey year (up to 3
longest)
spells not just the
and the dates of training periods by source. Potential sources of training included
business college, nurses programs, apprenticeships, vocational and technical institutes, barber
or beauty schools, correspondence courses and
training
company
programs are independent from training received
program which
is
included in the schooling variables.
tr ainin g.
All of the types of
a formal regular schooling
in
However, the questions ask about
only those spells of training that lasted at least 4 weeks (they did not have to be
This suggests that the
spells
NLSY
measure of training
than informal on-the-job training.
is
more
likely to capture
full time).
formal training
Therefore, tenure on the job will capture both
returns to seniority and returns to training programs lasting less than 4 weeks such as
informal on-the-job training.
For the analysis of the impact of training on wages a subsample of the 12,686
respondents has been selected.
(1280 respondents) and
all
I
have excluded the military subsample from the analysis
college graduates.
The
final
sample
is
composed of
individuals
who had completed their schooling by the 1980 interview date (where "completed" is defined
as not returning to school by the 1983 interview date).
requirement reduces the sample
respondents are
still
size substantially given the
17 or less in 1980).
The completion of
age
structvu-e of the
schooling
NLSY (4000
In addition, these individuals had to have
wage
observations at both the 1980 and 1983 interview dates.' This last restriction does not imply
that the respondent
had
to
be working
at the interview date since the
17
wage data used
are
wages in current or
last
These selection
job over the survey year.
sample of 3064 individuals that
will
and schooling for this subsample
training data are separated into three categories
- company
training (ON-JT),
firm (OFF-JT).
apprenticeships (APPT), and training obtained outside the
training obtained
of returns.
of training are allowed to have different types
possible to distinguish
during
OFF-JT includes
nurses programs,
from business courses, barber or beauty school,
courses.
vocational and technical institutes and correspondence
it is
final
and outcomes of training for non-college graduates.*
possible to examine the patterns
The
a
be used in the empirical work. Using a constructed
weekly event history of private sector training, employment,
it is
criteria yield
spells of training in
between
Each of these
three types
Since the data are longitudinal
each of these categories received
during current
employment with a previous employer and speUs received
employment In
addition, for training received
on the current
job,
it is
possible to identify
both completed and uncompleted spells of training.
presented.
In Table 2 characteristics of this sample are
training for this
-
14.7 percent
- who have
experienced
have had on-the-job training, and
and nonwhites who are
be kept
in
mind when
individuals in
"off-the-job" in
sample comes from
1.8
company
training
programs
also
The number of women
particularly small
and
the results in the next section.
seem small compared
found in other surveys such as the employer
are restricted to a speU of 4 weeks or
is
EOPP survey.
more of training,
18
-
only 4.2 percent of the sample
percent have been apprentices.
some of
may
terms of the percentage of the sample
this type of training;
in apprenticeship
interpreting
The primary source of formal
to
numbers
this
The number
that have
However, when the
as in the
NLSY,
needs to
of
been
EOPP data
the percentages are
remarkably
is
similar'.
The average
quite long.
weeks, and of
The average
ON-JT is 31
spell length of
an apprenticeship
is
63 weeks, of
OFF-JT
is
41
weeks. Finally, Table 2 shows that there are distinct differences
in the types of training received
IV.
length of time spent in these formal training programs
and the duration of
this training
by race and gender.
Private Sector Training and Wages: Empirical Results
Table 3 presents estimates of the probabilities of an individual receiving each of the
three types of training at
interesting patterns.
The
to the 1983 interview date as a function of their 1983
among
Differentiating
characteristics.
is
some time up
probability of investing in off-the-job training
male or has longer tenure on the job.^
the-job training
is
concentrated
experience^ but tenure in 1983
high
live in
more
imemployment
difficult to
these various types of training reveals
is
On the
other hand,
is
lower
the youth
company provided formal on-
among white married unionized males with
not significant. At the same time
areas.
if
some
it is
greater
work
lower for those
who
This suggests that as unemployment rises firms find
provide expensive formal on-the-job training to
new young
it
entrants.
in
an apprenticeship include being
white, unionized, and male. Interestingly, living in an high
unemployment area means you
Finally, the
are
more
by the
most important determinants for participating
likely to
fact that
have participated
in
an apprenticeship program. This may be explained
most apprenticeships are
in
construction and manufacturing which
experienced very high unemployment rates during
The
this period.
role of schooling in training decisions varies by type of training.
of non college graduates
of the equations
it is
when schooling
is
For
this
sample
included as years of completed schooling in each
never significant. However,
19
when the
schooling variable
is
broken into
the categories
- less
post high school but not
than high school high school graduate and
Staying on
coUege graduate- some different patterns emerge.
school degree or
some post high school experience
complete a high
significanUy increases the probability of
company provided on-the-
formal
receiving off-the-job training and (marginally)
school degree but
job training. Most apprentices have a high
who
in school to
it is
less likely that
someone
participate in an apprenticeship program."
has some post high school education will
The
fourth
column
in
in the 1983
Table 3 examines the probability of individuals
provided training in 1983 as a function of their
survey year to have participated in company
the
columns use characteristics in 1983 to predict
1983 characteristics. The previous three
by type (even prior to 1983).
probability of having ever received training
increases the
how
number of observations with
training
it
While
this
does not allow for the examination of
probabiUty of future training and the actual
previous speUs of training increase the
impact of current tenure on current
ON-JT
probabilities.
In column 2 tenure in 1983 was
the 1980of ever having received training during
not significant in explaining the probability
timing correctly,
significant. In contrast, by specifying the
83 period, whereas experience was
column four shows
receiving
that tenure with the current
ON-JT and
much more
those individuals
employer increases the probability of
who have had
training with a previous employer are
future.
Ukely to receive on-the-job training in the
Finally, I
occupation dummies
have also included broad industry and
training
for on-the-job training, off-the-job
results
and apprenticeship. Although the detailed
of space, a few
are not reported here for reasons
inclusion of industry and occupation
in the probits
dummies
20
summary comments are
did not change very
in order.'
much any
The
of the
coefficients for the
ON-JT and OFF-JT probits. However,
apprenticeship probit
factors
when
The
local
unemployment
rate
there are
some changes
and being male became
in the
insignificant
school graduate
industry and occupation were added but being white, a high
and a union member
still
raised the probability of participating in
expected, apprenticeships are
technical workers
and
in the construction industry
more common
craft workers.
sales workers, clerical staff or craft
an apprenticeship. As
For the ON-JT
probits, those
workers were more
likely to
employed as managers,
have experienced a speU of
wholesale and
formal company provided training while those in the
significanUy less likely to have received
this type
were
retail industry
ON-JT. For OFF-JT, there were two very
occupations that were more likely to have acquired
technical workers,
and among
of training
-
different
professional and
and service workers. None of the industry dummies were
significant for
off-the-job training.
Keeping these
in mind,
differential patterns in the acquisition of training
examine how these three types of training
affect the
wages of young workers are regressed on a
training,
and other
apprenticeships,
factors.
The
APPT. These
effects
spells of
fianction of tenure,
rate, the
work experience,
training variables are divided into
schooling,
OFF-JT, ON-JT and
variables are further separated into training received while
APPT and ON-JT
on current wages^°. The additional
unemployment
now
wages of non-coUege graduates. Log
employed with a previous employer and the current employer.
and uncompleted
I
number
Finally, I allow
completed
from the current employer to have
fartors in the
different
wage equation include the
local
the
of jobs held since finishing school, whether or not
respondent Uves in an urban area, marital
status, race, gender,
21
coverage by a collective
agreement, and health. Equation
1
in
Table 4 presents results from a standard log wage
dependent variable
equation specification excluding the training variables where the
log of wages in 1983.
is
the
Equations 2 and 3 in Table 4 include the training variables, with
contains the
equation 3 also adding broad industry and occupation categories. Equation 4
Heckman correction for sample selection. The sample selection issue will be
and
I first
focus
One
on the
results of equations 1-3 in
of the striking findings
is
Table
discussed later
4.
the insensitivity of the estimated coefficients on tenure
to the inclusion of the training variables.
It
appears the training and tenure are basicaUy
equations
uncorrelated since the coefficient on tenure does not alter between
tenure variable
it
may be
is
always significant in the wage equation although there are
capturing. Specifically, the training variables in the
speUs of formal training lasting at least one month but they
informal on-the-job training.
literature, tenure
may
In this case, the tenure variable
"tenure" effect plus the returns to informal training.
matching
NLSY
may
upwards (see Topel for a discussion of the
and
2.
many
The
factors
are good measures of
not capture
speUs of
all
capturing both a pure
is
In addition, as
represent job match quality so
1
its
size of this bias). Finally,
shown
in the job-
coefficient
biased
is
a positive tenure
effect
shirking and/or to lower turnover.
could reflect incentives provided by the firm to reduce
plays
Equations 2 and 3 in Table 4 show the significant role that training
determination.
Even
measures have a
after controlling for industry
significant
impact on wages.
and occupation the various
in
wage
training
Periods of off-the-job training and
raise wages significanUy.
apprenticeship training acquired before the current employer
Weeks
employer also raise wages.
of on-the-job training and apprenticeship with the current
22
Other variables that
living in
significantly raise
wages include
total
work experience, years of school,
Being
an urban area, male, white, married and coverage by a collective agreement
wages significantly.
disabled or living in an area with high local unemployment depresses
Adding industry and occupation dummies to the estimated wage equation
the size of the effect of training
on wages but the
training variables that
slightly
were
reduces
significant
they are added. Workers
without industry and occupation dummies remain significant when
employed
in the mining, construction
and transportation industries earn more
relative to
business, repair, personal
those in manufacturing while those in wholesale and retail trade,
and professional related services earn
less.
Professional, managerial
and
craft
workers
all
earn a wage premium relative to laborers and farmers.
In order to have a better sense of
relative to other factors such as tenure
wages
how
the different training variables affect wages
and schooling. Table 5 presents calculations of hourly
company provided on-the-job
training
raises
and apprenticeships,
training, especially
shows that
for different characteristics of the sample. This table
wages
substantially.
The
(keeping experience
impact of one more year of school or one more year of current tenure
the same) raises wages to almost to the
The
same amount
return to additional schooling and tenure
months of on-the-job
training
is
as 6
months of
off-the-job training.
even smaller relative to the return to 6
from the current employer. The
latter raises
wages by ahnost
raises wages by ahnost
ten percent while off-the-job training obtained before the current job
5 percent.
We
know from Table
3 that
women and
nonwhites are
receive on-the-job training. However, Table 5 shows that
if,
much
less likely to
for example, a nonwhite
male
earnings between himself and
obtains 6 months of off-the-job training he can cut the gap in
23
a white male with no training in
off-the-job training but the
findings
half.
White and nonwhite female wages
gap between female and male wages remains quite
on the role of training obtained from "for-profit" proprietary
for the current debate
ability of these institutions (see
to welfare recipients.
However,
this
Some
institutions.
weU
large.
with
These
institutions is important
on whether or not Graduate Student Loans and
be continued to be granted to students in these
concern about the
rise as
Pell grants should
have expressed
cities
INTERFACE (1989)) to provide training
paper shows that on average for
coUege graduates that off-the-job training from proprietary
this
institutions has
sample of non-
a sizeable impact
on wages.
Some
the variables that are
other interesting findings contained in Table 4 concern
training acquired before the current job
not significant. For example, spells of on-the-job
have no impact on current wages. This suggests that
to employer for
ON-JT
not portable from employer
because formal
young workers who are not coUege graduates. This may be
for these workers
trained workers
ON-JT is
is
more firm
specific than general.
who change employers are
training in the ftiture
may
also
not as able as those workers
job training but do not leave their employer".
that having received training
It
However, equation 4
from a previous employer
be because those
who receive
in
on-the-
Table 3 indicates
raises the probability of receiving
workers who
which does not seem consistent with considering trained
change jobs as lower quality workers.
Off-the-job training acquired before current
employment has a significant and positive
current employment
impact on wages, while off-the-job training during
This
may be because young workers who are acquiring training fi-om a
24
is
not significant.
proprietary instimtion
are planning to use this training to
findings
may
move
employer and career
reflect the sharing of costs of this training with the current
lower wages. In the following section
I
and employer mobility. Unfortunately,
is
to another
examine
in
more
employer through
between
detail the link
difficult to identify
it is
track, or the
from the
NLSY
training
data
who
paying for the direct costs of training received off-the-job.
Table 6 presents the findings using the specification of equation 3 in Table 4 but
broken down by various subsamples of
union
status.
some of
small.
It
interest according to gender, race, education
and
should be noted that given the sample composition, as shown in Table
the cell sizes (e.g. the
number of women
in apprenticeships)
become extremely
Nevertheless there are some interesting differences across these groups.
example, Johnson and
Youmans
potential impact of unions
(1970), Lewis (1986)
on wage
profiles
studies indicates that while unions raise the
union workers are
flatter
2,
For
and Mincer (1983) have discussed the
and job
wages of
training.
The evidence from many
members, the wage
their
profiles of
than that of their nonunion counterparts. The results presented
here confirm those findings. The union wage premium for the sample as a whole
20 percent yet the equations
in
Table 6 show that nonunion workers' wages
is
around
rise faster
during training spells than union workers' wages.
Another interesting finding is what happens to the
the sample
is
divided by educational level. While those
some post high school schooling
receive a
wage premium
coefficient
who have a
for
a high school degree actually receive lower wages during an
firms
may be
providing
more general
training for those
25
on current ON-JT when
high school degree or
ON-JT, those who do not have
ON-JT spell.
who do
This suggests that
not complete high school
costs of this training are shared
and the
receiving a lower
level
that the coefficients
is
for those
on
wage during the
between the workers and the firm with workers
Another finding related to educational
training period.
on race and gender become much
less significant
and smaller
continuing
post high school education- This seems to suggest that
who have some
gap in wages between males and females and nonwhites and
in school reduces the
whites.
Before reaching any
4-6
final conclusions
necessary to discuss in
it is
estimates due to self-selection.
in training
who
are
more
on the
basis of the results presented in Tables
detail the possible sources of bias in the training
As ah-eady mentioned, employers may only
programs who have some unobservable
more motivated would be more Ukely
place employees
characteristic, "trainability", or individuals
to pursue off-the-job training. In either case
measure will be biased upwards
the estimated coefficient on the various training
(i.e.
the
treatment selection problem).
A variety of ways
Heckman and Robb
to try to address this issue are described in
(1986).
One method
procedure using the probits in Table 3 for
Mills ratios as regressors in the
in equation 4 in
in the
Table
4.
This
that
I
used was a "standard"
ON-JT and OFF-JT with
wage equation. The
is
Heckman
results of this
Heckman
procedure are presented
if
the error terms
whether or not
two probit equations are not correlated. To examine
I
two-stage
the appropriate inverse
a relatively straightforward procedure
appropriate assumption for this sample
(1979) and
this
was an
estimated a bivariate probit for the probability of
(results available
receiving on-the-job training and off-the-job training
found the correlation coefficient to be small
26
(-0.12) with
a
t-statistic
upon
request) and
equal to
-1.67.
As
shown
in equation
4 in Table
4,
with the inclusion
none of the previous findings are altered
significant in the equation.
However,
form or somewhat
functional
common problem
a variable
tiiat
witii tiiis
identification using
A second
procedure for
approach to deal
tiiis
in botii
witii
can be expressed
f.
log (w,)
are
tiie
characteristics
characteristics in
f,
tiie
probits
and
tiie
it is
tiie
individual.
An
is
a
difficult to identify
wage equation."
self-selection varies
individual's
wage
= Z'.d +
+
fi
at
time
Ci.
which are individual
f,
The
tiie first
tiie
coefficients
results fi-om
column of
tiie-job training
tiiis
may be
time invariant
effects (botii
The
Fitting
differencing individuals'
observed and unobserved)
estimated witiiout bias.
presented in Table
second approach to sample selection are
results for tiie entire
and apprenticeships
ON-JT, however, are never
By
will lead to bias in estimates of b.
invariant
wages between 1983 and 1980, aU time
drop out, and
specific but
training.
correlated with whetiier workers undergo
may be
equation (1) while omitting
In
model since
type of
This
on
vary for each individual over time, and
a vector of variables affecting wages that
is
all
rests primarily
as:
(1)
where Z'
procedure
sample selection assumes that
for
only across individuals and not over time
t
tiiis
lambdas are not
exclusions of explanatory variables".
artificial
you would not include
that the
Note
regressors.
of the inverse Mills ratios (the lambdas) as
sample
it is
significantiy raise
significant for the entire
27
clear
wage
tiiat
additional
growtii.
weeks of
7.
off-
Additional weeks of
sample or any sub-group. This suggests
that there
also
may be some problem
be that the
models
in Tables 6
and
between training
effect
7.
those in a union job in 1983
distinguish
who have some ON-JT.
at
some
Moving
effect
is
The
only change
now a
is
that
weeks of off-the-job
The
significant factor.
spells across different
training for
specification does not
employers in the interval so there
institution that gets
them
may
into
later date.
to a job that
on the wage
may
remains similar between the cross section and fixed
be some workers who take a technical course in a proprietary
a union job
It
here are too small for a significant effect to be foimd. The size
cell sizes
and significance of the OFF-JT
effects
of selection bias for those
rate.
is
covered by a collective agreement has a large and positive
Those employed
in a
nonunion job in 1980 and a imion job
in 1983
experienced significant wage growth over the period, while those working in a union job in
1980 and a nonunion job in 1983 experienced a large decrease in their wage.
jobs^
at
Changing
any time during the 1980-1983 period increases wage growth for the sample as a
whole, but again there are differences across the various demographic groups. Only white
females and
all
education groups except those with some post high school education have
changes in their wage growth
much
if
they change employers.
Finally, tenure
on the job has a
larger return to nonunion employees than imion employees, as expected from the
earlier discussion
on imion wages
rV.
in
Table
6.
Private Sector Training and Wages:
While the returns
to a college degree or
Conclusions
government training programs
have been widely documented, there has been relatively
little
in the U.S.
analysis of the returns to other
forms of hiunan capital investment that non college graduates undertake. This paper has
28
shown
that private sector training plays a significant role in the detennination of wages and
wage growth of the 70 percent of young workers
college. Specifically,
when
private sector training
training, off-the-job training,
in the U.S.
is
who do
not graduate from
divided into different types (on-the-job
and apprenticeships) some very different patterns emerge. For
example, the characteristics that appear to influence the probability of receiving training are
primarily race and gender.
Women and
nonwhites are
much
less likely to receive training
within a firm either through an apprenticeship or other forms of on-the-job training. This
differential pattern in the acquisition of training
by race and gender
may be a
partial
explanation of the persistent wage gap between males and females and whites and
nonwhites.
Schooling
apprenticeships but
it
raises
the
probability
of receiving
off-the-job
training
and
had a smaller impact on the probability of receiving firm provided on-
the-job training.
All types of training raise wages significantly. In particular, this paper shows that on
average, for this sample of non-college graduates, off-the-job training from proprietary
institutions
can be useful for increasing wages. The impact of these training variables also
seems to be larger than the impact of tenure on wages. This paper does not argue that
there
is
no
role to be played by job matching or other explanations of rising
but rather that
is
when
there
is
wage
profiles,
appropriate data on training, the impact of training on wages
quite large relative to other factors for
young workers.
Finally, While on-the-job training with the current
employer increases wages with the
current employer, this type of training seems to be quite firm specific since on-the-job
training
from a previous employer
is
never significant for current wages. At the same time.
29
there seems to be
some evidence
workers on the job
it is
the-job training
primarily .specific
Institute
is
for those
that
general training
if
who have
is
is
being given to any group of
The
not completed high school.
finding that on-
consistent with recent findings fi'om the
Hudson
which surveyed 645 firms in the U.S. and foimd that only 8 percent had any
of general remedial on-the-job training programs".
The
fact that U.S. firms are
willing to invest in firm specific training than in general training
inability to "capture" the returns
on investments
is
sort
more
understandable given the
However, whether or
in general training.
not U.S. firms will be able to remain competitive with this strategy in the future, given the
characteristics of the
technology,
is
new
entrants into the workforce and the
demands of new
questionable.
Section V. Training and Mobility: Theoretical
The
skill
transition
from school
to
work
is
Framework and Data
typically a period in
which many young workers
experience a wide range of different jobs and experience some of their most rapid wage
growth over their working
individual's
life.
Hall (1982) has estimated that the
first
ten years of an
working career will include approximately two-thirds of all life-time job changes.
Topel and Ward (1988) found that over half of young male new entrants into the labor
more jobs
market held
six
male worker
in twenty
or
in their first ten years of
remained with
their first
work experience. Only one young
employer for ten years
in their sample. All
of this suggests that young workers' early years in the labor market involve several
employment
transitions.
The purpose of
early years of
this section of the report is to
examine for young workers
work experience the determinants of leaving an employer. In
30
in their
particular, this
on the
section focuses
role of different types of training
employer. In the previous section of
this rejwrt I
on the probability of leaving an
reached the following conclusions.
First,
formal company provided training, or ON-JT, appears to be highly firm specific in the U.S.
and, therefore,
raises
is
not portable from employer to employer.
wages in the current job but has no
employment Second, formal
OFF-JT, has
little
effect
effect
Company provided
on the wages earned
training
in subsequent
training received fi'om 'for-profit' proprietary institutions, or
on the wages earned on the current job but
it
does raise the
expected wage in subsequent employment. Finally, there are important differences by race,
gender and education level
and the impact
in the probability of receiving different types of formal training
this training
has on wages and wage growth.
These findings have several implications
implication
is
that
if
company provided
of leaving an employer should decline
training.
An
additional implication
programs they are more
training allows a
training
if
is
is
mobility.
a young worker has experienced some on-the-job
that
if
workers participate in off-the-job training
to change career paths
employer.
and find a
In this case, off-the-job
'better match'.
from the National Longitudinal Survey Youth cohort, NLSY,
this part
Using data
of the report
examines in detail the factors which influence the probability of new entrants leaving
first
One
primarily firm specific then the probability
likely to leave the current
young worker
on
for the impact of training
their
job including the differential effects of company provided training, apprenticeships and
training
from
'for-profit' proprietary institutions.
There are a variety of explanations of why young workers change
status so often in the early years of their careers
31
and then seem to
their
'settle
employment
down' into more
stable
where
seniority rules determine layoff policies,
at risk of being laid off in
a downturn. Even in the non-unionized
employment In the imionized
young workers are more
sector,
firms use seniority as a major determinant of
sector,
many
falling
demand.
whom
to lay off in a period of
There are other explanations of the higher turnover rates of young workers, however,
that have
little
to
do with the
state of
demand. The three main theoretical explanations
as detailed by
include job search, job matching and on-the-job training. Job search theory,
lippman and McCall
where
(1976), states that information about
to find
nature of that job are difficult to acquire, especially for younger workers.
accept employment and remain in that job as long as the
alternative wage. Therefore, workers
who earn more
wage paid
a job and the
Workers
will
in that job exceeds the
relative to their alternative
wage are
less likely to quit.
An
alternative explanation of turnover behavior can
1979b, 1984).
be found
in Jovanovic (1979a,
about the
In the Jovanovic learning model both workers and firms 'learn'
quality of the
unobserved characteristics of each other over time. As tenure increases, the
job match
is
discover the
revealed as firms observe workers' actual productivity and workers
non-pecuniary aspects of their job. In
relationship
'better'
this
model there are two countervailing forces
between tenure and the probability of leaving an employer.
On the
for the
one hand,
workers remain with employers longer leading to negative duration dependence
the probability of leaving a job.
On
in
the other hand, as 'bad' matches are revealed the
turnover probability will rise over time.
The process
of on-the-job training within the
32
human
capital
model as described by
Mincer (1974) implies that as workers acquire firm-spedfic
training, their productivity and,
Therefore, the probability of leaving an employer will
fall
with training and tenure since the wage will rise relative to the alternative wage.
In
consequently wages, will
addition, employers will
be
However,
in specific skills.
will
rise.
be either no
less likely to lay off those
if
most of the
on the
effect
workers in
initial training for
whom
they have invested
young workers
quit probability or the quit probability
All of these theories are not mutually exclusive and clearly
is
general, there
may even
rise.
some combination of
all
of these factors influences the probability of a young worker remaining with an employer.
Consequently,
not the purpose of this report to distinguish between these different
it is
theories. Rather,
it
would be more useful
if
precise data
on employment
spells
could be foimd in order to establish the links between different types of
and training
tr ainin g
and
turnover behavior.
There have been
the role of training,
employer. This
training
is
relatively
demand and
few empirical studies which have attempted to examine
other factors in predicting the probability of leaving an
primarily due to the lack of accurate data on the timing of private sector
and the lack of detailed employment
histories for workers.
Recent exceptions
include Gritz (1988) and Mincer (1988). Gritz uses data from the early years of the
NLSY
and finds that private sector training (not distinguishing between different sources of
training) increases the
amoimt of time
amoimt of time males were employed.
the
NLSY when
employment
in total
employment
Gritz's study uses data fi-om the very early years of
most of the observed training
history begins.
for females but decreases the
spells
occurred before the detailed
Mincer uses data on training and mobility from the Panel Smdy
33
of Income Dynamics, PSID. The training variable comes from the answer to the following
question in the 1976 and 1978 interviews:
average
new person
to
become
broad measure of training
it
fully trained
it
how
like yours
and qualified?" While
does not measiire
the specific respondent In addition,
previous
"On a job
how much
long does
it
take the
this is potentially
a very
training has actually occurred for
captures training information for the current job, not
employment
Using data from the
NLSY
it is
possible to examine in
possible in the past, the role of training, the general state of
characteristics in determining turnover.
whatever reason)
is
also
known
The
more
detail than has
been
demand, and other personal
probability of leaving
employment
as the hazard rate or failure rate in renewal theory.
(for
The
hazard rate or tvunover probability can be expressed as follows:
where
=
h(t)
(2)
g(t)dt
is
g(t)dt/(l
the probability of leaving an employer between time
the probability of being employed at time t and
employment. In
this
h(t;z)
h^Ct) is
t is
t
and t+dt,
=
is
G(t)
is
used:
zB
ho(t)e
an arbitrary and unspecified base-line hazard function and z
characteristics including training.
1 -
the duration of the current spell of
paper the following Cox proportional hazards model
(3)
where
G(t))
-
The Cox model
34
is
is
a vector of
convenient for dealing with right
censoring and
it is
nonparametric in the sense that
it
involves an unspecified base-line
hazard instead of making further distributional assiunptions such as those required for the
Weibull or Log-logistic hazard. However,
whether or not there
is
is
means
this
that
it
will
not be possible to measure
negative or positive duration dependence in employment, but this
not a key focus of this paper.
In a model of the role of training in the probability of leaving an employer
important to be able to allow
tr ainin g
it is
to occur over time with the employer. Allowing for
covariates such as training to be time dependent implies:
h(t;z(t))
(4)
where
z(t) is
a vector of
all
fixed
=
and time varying covariates.
Oakes (1984) the components of the vector
categories
of variables
-
ho(t)e^')«
treatments
z(t)
that
As
discussed in
Cox and
can be divided into the following three
vary
with
time;
intrinsic
properties
of
individuals/jobs that are time invariant; and exogenous time varying variables.
Obviously the different types of private sector training are the 'treatment' variables
of interest. Examples of time invariant personal and job characteristics include gender, race,
education, occupation, industry, union status, location of the job in an urban area, and
whether or not the respondent
purpose of
is
disabled.
this study include the local
Time
varying 'exogenous' variables for the
unemployment
rate, marital status
and the nvmiber
of children.
For the
analysis presented in this part of the report a different
35
sample
is
used to
analyze mobility patterns than was used to examine the determinants of wages. This sample
uses
more recent
years of the
NLSY. As
wage
in the
analysis I have excluded the 1280
respondents in the military subsample from the analysis. However,
respondent
who
have also deleted any
has completed school before the 1979 interview year. The final sample
a pooled sample of yoimg workers
who have
of school leavers
~
those
who
left in
first
The estimated hazard models the determinants of
after leaving school
permanently for
this
it
wage
However,
I
substantially reduces the
is
made up
and 1983.
of 5 waves
In addition, the
year after 'permanently exiting school.
the turnover probability for the
This sample has
sample.
given the age structure of the
graduates in
sample
1979, 1980, 1981, 1982
respondents had to have obtained a job in the
is
school and not returned to school for at
left
least four years ('permanently' out of school). Therefore, this
study.
I
NLSY
compared
many more
first
job
college
to the sample used for the
do not include anyone who completed school before 1979, which
sample
size.
In addition,
to leave school over the period (1979-1983).
I
do not attempt
Obviously
this
to
model the decision
was a period
in
which many
young people may have delayed entry into the labor market given the high unemployment
rate.
I
include
dummy variables
for year of entry in the following analysis but future
work
would benefit from a complete modeling of the schooling/employment/training decisions
taken by young workers.
Characteristics of this sample are presented in Table
almost three quarters of the sample leave their
after school.
years)
is
first
The average duration of employment
about a year and a
half.
8.
As can be seen
employer during the
(including those
still
first
in
Table
8,
four years
employed
after four
Almost seventeen percent of the sample experienced some
36
form of formal training during
their first job but the distribution of this job training
source varied substantially by demographic group. College graduates were
to have received
more
likely to
some form of ON-JT while those with
ON-JT by gender (not controlling for other factors).
some of the
cell sizes for tr ainin g
in
mind when
likely
important to note that
It is
by demographic group are extremely small and
interpreting
some of the following
than
difference in the probability
of receiving
be kept
diploma were
Women were more
men to have received some form of OFF-JT but there was little
to
much more likely
just a high school
have participated in some form of OFF-JT.
by
this
needs
results.
Table 9 presents more detailed information on the relationship between tenure on
the job with the
first
employer and the various types of training. The
over 80 percent of the sample have
market. Those
who
left their
any formal ON-J-T (only 13
employer
%)
more (8.1%). The pattern is a bit
a quarter of those who
this
left their first
who
were much
stayed with their
different with participation in
left their first
panel shows that
employer by the fourth year in the labor
relatively early
than those
first
less likely to
first
have had
employer 3 years of
OFF-J-T programs. Almost
job between 2-3 years received OFF-J-T. However,
percentage drops dramatically for those with 3 or more years on the job to only 11.7
percent.
The second panel
on having participated
in
is
perhaps even more interesting. This panel shows, conditional
one of the types of private
during the tenure with the employer.
wage determination one view of
As
training,
that tr aining spell begin
discussed in the previous sections
a
'test'
other words, firms use formal training programs as a
way
training
when
is
that
37
it is
(Weiss and
on
Wang
training and
(1990)).
In
to avail themselves of private
information
who
known only by the workers. Workers who fail
pass do not leave.
This would imply that
we
in
a workers's tenure with the firm. However, in
of
ON-J-T
spells begin after
first
costiy formal
spells that last
this
second panel
at the firm.
is
a match, and
if yes,
ON-J-T. Since the measure of training used
4 weeks
it
may be
we
see that 60 percent
This seems to be more
where firms(workers) make a determination within the
6-12 months on whether or not there
more
should observe ON-J-T occurring early
one year on the job
consistent with a job matching story
the test leave the firms and those
the firm then invests in
in this
paper only captures
possible that shorter formal or informal training spells are
used early in the career with an employer as an indication of match quality and longer
training spells follow later.
Contrary to the timing of ON-J-T spells almost 60 percent of spells of OFP-J-T begin
within the
first
year with an employer. This
to obtain training that they
need
may be due
to
for their current job, or
employees going outside the firm
employees deciding that there
is
not a job match and seeking a training program that will allow them to leave their current
employer and get a better job. Finally and not
surprisingly,
most apprenticeships begin very
early in the tenure with an employer.
VI.
The
Private Sector Training and Mobility: Empirical Results
results obtained
from estimating the Cox proportional hazard with time varying
covariates are presented in Tables 10 and 11.
an
asterisk-
equation
1,
The time
that
The time
varying covariates are indicted by
invariant intrinsic characteristics of the individuals /jobs in Table 10,
seemed
to influence the probability of leaving
disabled, union status, race,
and school
level.
38
an employer included being
Disabled respondents were more
likely to
leave their employer while being employed in a job covered by a collective agreement or
being a college graduate significantly lowered the probability of leaving the
Blacks were
hispanics.
more
likely to
There was no
have shorter durations on their
significant effect
first
employer.
first
job than whites and
on the length of time with the
employer by
first
gender. However, there were significant differences in expected length of employment by
school attainment. Those with a high school degree or less were
more
likely to leave their
employer, whereas those with a college degree were less likely to leave.
Of
the time varying 'exogenous' covariates the local
significant implying that those
leave their employer.
who
The hurdle
lived in high
significant effect
With regards
JT were much
of
likely to
less likely to leave their
training
OFF-JT
more
be
to have
no
job. Finally, those workers
who were
young people who had some formal ONin
some form
ON-JT
is
more
'general'.
This seems to suggest that
These findings are consistent with the
firm
results
on
and wages.
In equation 2 the hazard
The
seemed
to
employer while those who participated
likely to leave.
is
less likely to
their first employer.
to the training variables, those
OFF-JT were more
specific while
remain with
first
children
was
rate
imemployment areas seems
The number of
on the expected duration of the
married were more
unemployment areas were
for youths in high
getting a job rather than keeping one.
unemployment
inclusion of industry
is
re-estimated including industry and occupation dummies.
and occupation does not change the
the variables in equation
1
young workers employed
in construrtion, wholesale
coefficients or significance of
with the exception of college which becomes insignificant. Those
39
and
retail,
and business,
repair, personal
and professional services were much more
manufacturing.
The
employers than those in
likely to leave their
with managers
only significant occupation was managers
more
likely
to remain with their first employer.
In equation 3 of Table 10 an additional variable
is
added which
is
the difference
and a log predicted wage. The
between the log of the current wage (which varies with time)
wage
predicted
is
estimated coefficients
obtained by the formula in Table 8 which uses the
from a log wage equation for the
starting
wage
for this sample.
Those individuals who are
wage are more Ukely to leave their employer
being paid less than their predicted alternative
as
in both equations 3
shown
equations
1
and 4 of Table
None
of the previous findings from
and 2 are altered very much.
In Table 11 the proportional hazard
of interest.
10.
Now
examine. Again,
is
re-estimated for various demographic groups
the results change dramatically depending
it is
upon which sub-group you
important to remember that some of the ceU sizes
now
results in Table 11. Nevertheless,
so care must be taken in interpreting the
to see
how
the results from the previous table change
demographic categories of
interest.
when
the sample
are very small
it is
is
interesting
divided into
high
For example, males, females, and blacks who are
duration on the
school dropouts have a shorter expected
first
job after they leave school.
has no
However, being a male or black high school graduate
effect
on the duration of
lowers the duration of employment.
employment, while being a female high school graduate
expected durations of employment, while
Male and black coUege graduates have longer
there
is
no
effect of a college degree
on the probability of females remaining with
employer.
40
their first
The
regressors
differences by race
and the
and gender are even starker when one examine time varying
effect of training.
For women, having additional children
lowers the expected duration of their
additional children.
At the same
first
job relative to those
time, there
is
no
an employer
Finally,
for males
and
women
ON-JT and OFF-JT
employment
for males
employment
in the first job for
When
but there
now
are
is
emerge. For example, those
no
still
not have
on the expected
lowers the probability of leaving
effect of marital status for blacks.
insignificant determinants
of the duration of
However, ON-JT increases the length of time
and blacks.
the sample
is
women who do
effect of children
duration of male or black employment. Being married
significantly
women
while
OFF-JT
in
increases their turnover probability.
divided by educational attainment other interesting results
who
are high school graduates or had
some post high
school
education and are covered by a collective agreement are less likely to leave their employer.
For the sample as a whole there
is
no difference
in the probability of leaving
between males and females. However, when the sample
males are
less likely to leave their
school degree or a college degree, but they are
post high school education.
employer only
if
more
divided by educational level,
is
employer than females
an employer
if
they have less than a high
likely to leave if they
have had some
In addition, being black raises the probability of leaving an
the young worker had a high school degree but was not significant for any
of the other educational groups.
The number of
children seems to affect the duration of
employer only for high school graduates, while marital status
graduates and high school dropouts.
In addition, the
41
is
employment with the
first
significant only for college
unemployment
rate
is
now
only
significant for high school graduates.
probability
if
Finally,
ON-JT
the respondent had a high school degree or
this probability for those
appears to lower the turnover
less,
while
OFF-JT seems
to raise
with a high school degree. Given the small cell sizes one must be
cautious in drawing conclusions on variables that are insignificant, but the different effects
of variables of interest by race and gender are quite striking.
VII.
Private Sector Training and Mobility:
Conclusions
This section of the report has focused on the link between training and the
probability of leaving an employer.
A high
percentage of ON-J-T spells begin after yoimg
workers have remained with their employer for at least one year.
consistent with a job matching story
6-12 months
first
more
where firms(workers) make a determination within the
on whether or not there
costly formal
is
a match, and
to
job, or
employees going outside the firm
employees deciding that there
will allow
them
Overall there
is
yes, the firm
is
first
then invests in
need for
However, when the sample
is
almost
may be
their current
not a job match and seeking a training program that
and get a better job.
significant differences in the patterns of job mobility
no difference
spells,
year with an employer. This
to obtain training that they
to leave their current employer
There are
if
ON-J-T. In contrast to the pattern associated with ON-J-T
60 percent of spells of OFF-J-T begin within the
due
This seems to be
in the probability of leaving
by race and gender.
an employer by gender.
divided by race, gender, and educational attainment there are
important differences between males and females. For example, children appear to have
little
affect
on the probability of males leaving an employer. At the same time, they have
a significant and positive effect on the probability of
42
women
not remaining with their
Among high
employer.
men
school dropouts and college graduates
to have shorter spells their
school graduates. In contrast,
men
are
more
job, but there
first
among
those
is
young workers in the U.S. appeared
more
The
general.
off-the-job
more
training being
to
be
qiiite
firm specific whereas off-the-job
results presented in
likely to leave their
more
than
this report indicated that on-the-job
Tables 10 and 11 seem to
likely to
remain longer
with their employer which would be consistent with firm specific training. Those
off-the-job training are
likely
post high school education
Those with on-the-job training are more
reinforce this conclusion.
more
employer.
likely to leave their
training appeared
are
no gender difference among high
who have had some
Evidence presented in the previous sections of
training for
women
this
we see
obtain
would be consistent with
However, when the sample
general.
race,gender and educational attainment
employer and
who
is
divided by
that the training variables are only significant
in the equation for females.
Overall
appears that blacks are more likely to leave their employer but
be true for those blacks who received
to only
seem
it
to
be any
there does
to
appears
a high school diploma. There does not
significant difference in the results for hispanics relative to whites.
seem
Finally,
be some evidence that blacks who receive some on-the-job training have
longer expected job durations in their
While
just
this
this part of the report
first
job.
has attempted to shed
new
light
on the
skill
formation
process of young workers and the consequences of this on their patterns of mobility there
are
still
many
issues that
the duration of the
first
remain unresolved. This report has modeled the determinants of
job after school, not subsequent employment.
43
As
the
NLSY
age
future research should examine
change over time.
industry
It
how some
would also be
and occupational
interesting to
categories. Finally,
findings are after additional
Nevertheless, there
of the gender, race, and educational differences
is
work
is
done
it
examine the hazard rates by broad
would be important to see how robust the
to address the endogeneity issue for trainin g,
a stoiy that emerges from the results in
workers and private sector training. Company training in the U.S.
for
yoimg workers
in their first job.
Yoimg workers
is
this report for
very firm specific, even
entering the labor market can receive
both 'good' and 'bad' draws from the labor market. There are some workers
draw who appear
in 'good' jobs are
to
move
more
to better
employment by
spite of the fact that
women
which
results in higher
effects are particularly strong for
are less likely to receive on-the-job training.
44
who get a 'bad'
investing in off-the-job training.
likely to obtain on-the-job training
a lower probability of leaving the firm. These
young
Those
wages and
women
in
FOOTNOTES
*
^
CEO
Kerns, David,
Kids", Industry
Even
Week. July
if
Xerox Corp.
in
W.
Miller,
"Employers Wrestle with Ehimb
4, 1988.
the training
is
entirely firm specific
you might
still
expect in
some
cases to
observe a positive effect of past
on wages with a future employer because if the
employer providing the training gives some wage premiimi for specific training (to lower
turnover) then the worker's reservation wage should be higher. The size of this effect,
however, becomes an empirical question.
tr aining
^
Given the age structure of the sample the restriction that the respondent had to
have completed all schooling by the 1983 interview date substantially reduces the sample.
The restriction of having wage data further reduced the sample size but this effect was not
as large. In addition, only about 120 respondents had completed college by 1980 who also
had wage data. Therefore, future work with the next waves of the NLSY might examine the
role of private sector training for college graduates.
The data for the training variables come from the starting and ending dates of spells
of training by source. These dates are given by month and year. In order to match this to
^
employment and schooling histories I assiune that all training commences and
ends at the beginning of the month. In the case of a spell which has the same beginning
and ending month I make the ending week the first week of the following month. If many
spells of training were quite short in duration this approximation might be inappropriate.
However, since all training spells have to be at least 4 weeks and the fact that the average
duration of training for this sample is around six months this should not be too serious a
problem.
the weekly
Dan
Black for very kindly ruiming the comparable numbers
for the EOPP data. The EOPP data are of hours of training rather than weeks, however,
he found that 3 percent of the EOPP sample had training of over 100 hours (one might
assume 4 weeks of 25 hours per week) and 2 percent had training over 140 hours (4 weeks
of 35 hours). The NLSY number for those in firm provided training lasting at least 4 weeks
^ I
is
42
would
like to
thank
percent
Tenure is specified as total weeks on the current job. In an alternative specification
was
represented by a series of dummy variables: less than 6 months; 6 months - 1
tenure
year; 1-2 years; and greater than 2 years. This had little impact on the fin din gs presented
'
here.
school
'
Experience
*
The
is
is
total
number of weeks of work
since finishing school.
assume that the decision about when to finish
exogenous with respect to decisions about post-schooUng training. However, given
probits presented in Table 3
45
more than 70 percent of U.S. youths do not finish college it is interesting to examine,
conditional on completing school, how and who of the non-coUege graduates acquire training
after school.
Future work should examine a more complete model of human capital
accumulation from school to training and government training programs.
that
The
'
used and the percent in each (in ()) are: agriculture,
mining (13); construction (5.6); manufactiuing (20.0) (omitted
category); transport, commercial and public utilities (5.0); wholesale and retail (27.1);
finance, real estate and insurance (5.4); business and repair services (5.9); personal services
The
(6.9); professional and related services (152); and public administration (43).
occupation categories include: professional and technical (62); managers (3.8); sales (5.5);
clerical (242); craft workers (103); operatives (182); laborers and fanners (10.7) (omitted
category); service workers including private household (21.1). Detailed results are available
firom the author upon request.
forestry
and
*®
industrial categories
fisheries (33);
The
on completed and uncompleted OFF-JT were never
wage equation specifications.
coefficients
different in any of the
significantly
you view training of young workers as a "test" as discussed in Weiss and Wang
this would be consistent with an argument that formal training programs are
a method firms use to avail themselves of private information known by workers. Workers
who "fail" the test leave the firm and those who "pass" do not leave. This discussion suggests
that it would be important in future work to also examine the mobility patterns of these
workers and the role of different types of training in the mobility pattern.
*^
If
(1990), then
^ The
probits used from Table 3 were
ON-JT and column 1 for OFFvariables included in the ON-JT probit
column 2
for
The only differences between the explanatory
and the wage equation are that education is entered as a series of dummy variables in the
probit and as years of completed school in the wage equation. Industry and occupation
JT.
dummies are only included
^ Another
in the
wage equation.
use instrumental
variables and include the conditional expectation of weeks of training in the wage equation.
To do this I first estimated individual probits for each of the types of training (separated
into training
strategy to deal with selection that
is
less restrictive is to
from a previous employer and current employer).
I
then estimated, using OLS,
separate equations, conditional on having experienced each of the types of training (also
separated into current or previous employer), where the dependent variable was the nimiber
of weeks of trainin g I then created an expected value of weeks training by type for each
observation using the probits and the OLS estimated coefficients and re-estimated the wage
equation using I.V. Tlie results (available from the author on request) are not reported for
the sake of brevity but again they suggest that the conclusions reached above are not
altered.
**
In other words, there
may be some young workers who
46
take a technical course in
a proprietary institution that gets them into a union job at some later date. In
coefiQcient on OFF-JT might become positive and significant
^ These union dummies do
some
not
come out from
strictly
this case the
differencing but they reveal
interesting patterns.
^'
The change job
variable
equation. In the cross section
fixed effects
I
included simply a
I
is
specified differently than
included the
number
dummy variable
it
was
in the ctoss section
of jobs since fini s hing school. In the
equal to
1 if
the respondent changed jobs
at all in the 80-83 interval.
" from
the
New York
Times. "Shortage of Skilled Workers
Fowler, July 31, 1990, p. D16.
47
is
Expected", by E.
REFERENCES
Abraham, K. and Farber, H. (1987). "Job Duration,
Economic Review. Jime, pp. 278-97.
Altonji,
Seniority,
and Earnings," American
Joseph and Shakotko, Robert (1987). "Do Wages Rise with Job Seniority?". Review
Economic Smdies. July, pp. 437-59.
of
Barron,
Black, D. and Loewenstein,
J.,
M.
(1987).
"Employer Size: The implications for
and wage growth", Journal of
search, training, capital investment, starting wages,
Labor Economics.
Jan., pp. 76-89.
Barron, J., Black, D. and Loewenstein, M. (1988). "Job Matching and On-the-Job Training",
Journal of Labor Economics Jan., pp. 1-19.
.
Ann
Bartel,
(1989).
Productivity:
"Formal Employee Training Programs and Their Impact on Labor
Evidence from a human resource survey", NBER working paper no.
3026.
Blackburn, M., Bloom, D. and Freeman, R. (1989). "The Declining Economic Position of
Less-Skilled American Males", NBER working paper no. 3186.
"Are Those Paid More Really No More Productive? Measuring the
relative importance of tenure as on-the-job training in explaining wage growth",
Brown,
J.
(1983).
Princeton Industrial Relations papers.
Brown,
J.
(1989).
"Why Do Wages
Increase with Tenure?". American
Economic Review
.
Dec, pp. 971-991.
"The Learning Enterprise: A Report on the Size and Scope of
Training", Training and Development Journal pp. 18-26.
Camevale, A. (1986).
,
Cox, D.R. and Oakes, D., Analysis of Survival Data. London:
Chapman and
Hall, 1984.
Mark, "The Impact of Training on the Frequency and Duration of Employment",
mimeo. University of Washington, 1988.
Gritz, R.
Hall, Robert, "The Importance of Lifetime Jobs in the U.S.
Review. September 1982,
Heckman,
J.
Economy". American Economic
72, 716-24.
(l979). "Sample Selection Bias as Specification Error". Econometrica. pp. 153-
161.
48
Heckman,
J.
and Hotz,
Among Alternative
(1989) "Choosing
J.
for Estimating the
Impact of Social Programs:
NBER working paper
The
Nonexperimental Methods
manpower
case of
training",
no. 2861.
Heckman, J. and Robb, R. (1986). "Alternative Identifying Assumptions in Econometric
Models of Selection Bias". Advances in Econometrics vol. 5, JAI Press, pp. 243-287.
,
INTERFACE
Unfair at
(1989).
Anv
Price:
Wel fare
recipients at
New York
Proprietary
Schools, report, N.Y., N.Y..
Jovanovic, Boyan, "Job Matching and the Theory of Turnover", Journal of Political
Economy October 1979a, 87, 972-90.
.
Jovanovic, Boyan, "Firm-Specific Capital and Turnover", Journal of Political Economy.
December 1979b, 87, 1246-60.
Jovanovic, Boyan, "Matching, Turnover, and Unemployment". Journal of Political
Economy
.
February 1984, 92, 108-22.
Katz, Lawrence and Revenga,
versus Japzm",
Ana
(1989). "Changes in the Structure of
NBER working paper
Wages: The U.S.
no. 3021.
Lalonde, R. (1986) "Evaluating the Econometric Evaluations of Training Programs with
Experimental Data", American Economic Review pp. 604-620.
,
Lang, K. and Ruud, P. (1986). "Returns to Schooling, Implicit Discount Rates and Black-
White Wage
The Review of Economics and
Differentials".
Statistics . Feb., pp. 41-7.
"Agency, Earnings Profiles, Productivity and Hours Restriction",
American Economic Review Sept. pp. 606-620.
Lazear, E. (1981).
.
"The Narrowing of Black- White Differentials
Lazear, E. (1979).
Economic Review
Leigh, D.
(1989)
Dlusionary". American
of Training 'Work' for Noncollege
Bound Youth? paper
,
GAO.
and Tan, H.
effects?".
is
Sept. pp. 606-620.
What Kinds
prepared for
Lillard, L.
.
(1986).
"Private Sector Training:
Who
gets
it
and what are
its
Rand monograph R-3331-DOL/RC.
lippman, Steven and McCall, John, "The Economics of Job Search:
Economic Inquiry June 1976, 14, 155-89.
A
Survey, Part
I",
.
Lynch, Lisa
M. (1991a) The Role
of Off-the Job
49
vs.
On-the-Job Training for the Mobility
of
Women
Lynch, Lisa
Workers", American Economic Review.
M. (1991b)
"Private Sector Training
May
1991, 81, pp. 151-56.
and the Earnings of Young Workers",
American Economic Review, forthcoming.
W.
Miller,
Mincer,
J.
(1988).
"Employers Wrestle with
Dumb
(1989). "Job Training: Costs, returns
Kids", Industry
and Wage
Week.
Profiles",
July 4.
NBER working paper
No. 3208, Dec. 1989.
Mincer,
(1988).
J.
"Job Training,
Wage Growth and Labor
Turnover",
NBER
working
paper no. 2090, August.
"Union Effects: Wages, Turnover, and Job Training",
Labor Economics pp. 217-252.
Mincer,
J.
(1983).
in
Research in
,
Mincer,
J.
(1974).
Schooling. Experience and Earnings
.
New
York: Columbia University
Press.
M. and Shack-Marquez, J. (1986). "Earnings and Different Types of
mimeo, BLS and Board of Governors of the Federal Reserve.
Pergamit,
Rothschild,
M. and
Stiglitz, J.
Training",
(1976). "Equilibrium in Competitive Insurance Markets:
An
essay on the economics of imperfect information", Quarterly Journal of Economics.
pp. 629-649.
Salop,
and Salop, S. (1976). "Self-Selection and Turnover
Journal of Economics pp. 619-627.
J.
in the
Labor Market". Ouarteriv
,
Stiglitz,
J.
"Incentives,
(1975).
hierarchy". Bell Journal of
Topel, R.
(1987).
Risk and Information:
Economics pp. 552-579.
Notes towards a theory of
,
"Wages Rise with Job
Seniority",
M.
mimeo. University of Chicago.
(1988). "Job Mobility and the Careers of
2649.
working paper no.
Topel, R. and Ward,
Weiss,
NBER
"A Sorting Model of Labor Contracts:
Layoffs and Wage-Tenure Profiles", mimeo, Boston
Andrew and Wang, Ruqu.
ImpUcations for
Young Men",
(1990)
University, June.
50
TABLE
1 -
EXAMPLES OF TRAINING QUESTIONS
Data: Panel Study of Income Dynamics, 1976-1980
"On a job
like yours,
how
long woiild
it
take the average person to
become
fully
qualified?"
"Are you learning
promotion?"
skills
on the current job which could lead
National Longitudinal Survey, Young
Cohorts
"Do you
&
to a better job or
Older Mens and Young
Women
receive or use additional training (other than schooling training) on your
job?"
"What was the longest type of
training
you have had since the
last interview?"
Current Population Survey, January 1983
"What
to
was needed to get the current or
improve skills on the current job?"
training
Employment Opportunity
Pilot Project Survey,
last
job and what training
EOPP
-
is
needed
Individual Survey
"Describe up to 4 training events occurring between 1/1/79 and the interview data
in 1980" (approx. 1 1/2 years)
EOPP
-
Employer Survey
"Number of hours typically spent by a new employee in the position last filled
watching other people doing the job rather than doing it himself during the first 3
months of employment"
"Number of hours a new employee
in the position
National Longitudinal Survey Youth Cohort,
spends in formal training"
NLSY
your schooling, military and government-sponsored training programs,
did you receive any other types of training for more than one month?"
"In addition to
"Which category best describes where you received this training"
(Both questions asked for up to 3 training spells per year)
51
TABLE
2
TABLE
Variable
Constant
3 -
PROBITS FOR THE PROBABILITY OP RECEIVING TRAINING BY TYPE
T-Stati«tic« in ()
Of£-th«-Job On-the-Job
Probit
Probit
Apprantica On-tha-Job
Probit
in 1983
TABLE 4 - DETERMINANTS OP LOG WAGES AT 1983 INTERVIEW DATE (N-3064)
T-Statistics in ()
Varlabl*
TABX^
Ca«« 1.)
Caae 2.)
Case 3.)
Case 4.)
5
- PREDICTED HOURLY WAGES BY SELECTED
CHARACTERISTICS
Whit* mala, average charactariaticst
no training
24 wka pravioua OFF-JT
24 wka complatad current ON-JT
24 wka pravioua apprenticaahip
24 wka cooplatad current apprenticaahip
1 additional year of achool
1 additional year of tenure
$5.47
5.74
5.96
6.17
5.74
5.64
5.65
Nonwhite male, average characteriaticas
no training
24 wka previoua OFF-JT
24 wka completed current ON-JT
24 wks previous apprenticeship
24 wks completed current apprenticeship
1 additional year of achool
1 additional year of tenure
$5.00
5.24
5.45
5.64
5.25
5.16
5.18
White female, average characteristics:
no training
24 wks previous OFF-JT
24 wks completed current ON-JT
24 wks previous apprenticeship
24 wks completed current apprenticeship
1 additional year of school
1 additional year of tenure
$4.71
4.94
5.14
5.31
4.94
4.85
4.88
Nonwhite female, average characteristics:
no training
24 wks previous OFF-JT
24 wks completed current ON-JT
24 wks previous apprenticeship
24 wks completed current apprenticeship
1 additional year of school
1 additional year of tenure
$4.34
4.56
4.74
4.90
4.56
4.48
4.49
*using the estimated coefficients from equation 2 in Table 4. Average
characteristics are: single, high school graduate, 99 weeks of tenure on the
job, 193 weeks of work experience, local unemployment rate of 10.01%, living
in the inner city, healthy, not covered by a collective agreement, and 2 jobs
since finishing school.
55
TABLE 6 - DETERMINANTS OF LOG WAGES AT 1983 INTERVIEW DATE BY
TABLE 6 CONTINUED
Varlmbl*
TABLE 7 - FIXED EFFECTS ESTIMATES
Dependent Variable: Log Wage (83) - Log Wage (80)
Variable
All
Constant
il
Experience
(wks)
^ Tenure
(wks)
^ON-JT
(wkB)
^OFF-J-T
A
(wks)
APPT (wks)
Job change dummy
Union83-Union80
R squared
Sample Size
TABLE
Varlabl*
7
CONTINUED
ZABLB 8 - SAMPL8 CBARACTBRZSTZCS (ll>2522)
Varlablai
XABLB • (continuad)
Xndustrv
TABXiB 9 - C3iaraet*riBtles of PrlTSt* Saetor
Coapl«t*d X«nur« by
Conpl«t*d T«nur«
%
Training
with Xraining by Typ«
XAK*
OF IXAVIHO
10 - DETERMZHAIITt OP THB PROBMIILITT
Variabl*
Kq. 1
Urban
# Children*
Disablad
lUrriad*
Union
Black
Hispanic
Male
!«•• than H.S.
High School
College
Medium Orate*
High Urate*
ON-JT*
OFF-JT*
Apprentice*
Kq. 3
Kq. 2
-.06
-.06
-.04
(-1.31)
.09
(1.59)
.22
(1.98)
-.22
(-3.44)
(-1.25)
.09
(1.49)
.24
(2.10)
-.22
(-3.41)
-.22
(-3.32)
.11
(1.98)
.09
(1.36)
(-0.95)
.09
(1.54)
-.28
(-4.34)
.14
(2.41)
.05
(0.83)
-.05
(-1.20)
.69
(8.51)
.26
(4.16)
-.24
(-2.79)
-.17
(-2.95)
-.17
(-2.74)
-.40
(-2.62)
.10
(1.51)
.03
(0.13)
.22
(1.96)
Log Likelihood
no
no
-14697.7
Kq. 4
-.04
(-0.85)
.08
(1.33)
.24
(2.10)
-.21
-.21
(-3.27)
-.27
(-4.16)
.11
(1.99)
.04
(0.60)
(-3.25)
-.21
(-3.21)
.09
(1.50)
.07
(1.19)
-.06
-.06
-.06
(-1.19)
.61
(7.32)
(-1.32)
(-1.17)
.58
(6.95)
.18
(2.89)
-.11
(-1.11)
-.17
(-2.88)
-.17
(-2.71)
-.22
(-1.46)
.10
(1.49)
.67
(8.22)
.23
(3.65)
.23
(3.62)
-.24
(-2.73)
-.13
(-1.35)
-.16
(-2.74)
-.17
(-2.70)
-.30
(-1.98)
-.18
(-3.13)
-.17
(-2.83)
-.32
(-2.12)
.11
(1.70)
.10
(0.48)
.09
(1.40)
.08
(0.40)
-.64
(-10.04)
Log Wage Diff*
InduBtry 6
Occupation dummiee
BiPWTBl
no
no
yes
yes
-14640.7
Notes:
63
-14644.4
.14
(0.67)
-.64
(-10.06)
yes
yes
-14592.0
ZABLB 11 - DBTXXMXlOarES OF ZHB PROBABZLZTT OP LKAVZMO BMPLOTER
BY DBNOORAPBZC OROUP
Variabl*
Urban
# Chlldran*
Disablad
Mairried*
Union
Black
Bispanic
Male
LttSB
than M.S.
High School
College
Medium Urate*
High Urate*
ON-JT*
OFF-JT*
Apprentice*
Log Wage Diff*
Log Likelihood
Number of Obs.
Males
-.06
Females
Blacks
XABLB 11 - DBTBXNZMJUfTS OF THB PROBABXLITT OP LBAVZMO BIPIiOTBR
BY OBMOORAPBZC ORODP (continuad)
Varlabl*
Post B.8.
Coll*g«
2876
V
135
Date Due
AUG.
3
1
19J5
Lib-26-67
MIT LIBRARIES
3
=1060
007niD0
?