me National hngitudtial Sueys (NLS) pmgrm supporrs mmy studies designed
m tiermdastmdrng of h. labor m%ket md mtiods of data collection. R. Dis.
cusion Pupers saies hw~rat% some of tie resach comtig OU1of the progrm.
Mmy of tie pa~m k his saics have bmn delivered ~ find rqorts .o_issioned
u
pzt of m exummd
pmt progra.
mder wtich gms
~c awaded though a
com~tiuve proc~s. ~e series is tile”ded 10circtiate tic fmdi”gs LO klcreskd reders
witi md o“tsi& tie Buea” of Lakr Smtis~ics.
P-ins htffes~d in obtatimg a copy of my Disc&ion Paper, othm NLS repr~, or
gmerd kfomation about tie NM Fogm
ad sweys should contact be Bweau of
Lakr Statitics, Office of Economic Rcsemch, Wmhtig~o”, DC 20212.0001 (202-6067m5).
Matid
in WISpublication is ti tie public domin md, with appropriateadi[, may
b. repmd”ced Wi&OU[ p-ksioa
@tiom
md conclusions ex~essti in his document z. hose of tie autior(s)
ad do not ne.=s%ily r~resml m o~cid position of tie Bucau of L* I S1alistim
or tie U.S. Depmtmcnt of Labor.
.
.
NLS
Discussion
Papers
.
.
Gender Differences in the Quit Behavior
of Young Workers
.
Audrey Ught and Manuelita
Ureta
Unicon Research
Corporation
March
.
1990
.
.
~s
find paper was funded by he U.S. Depatient
of Labor, Bureau of Lahr Statistics under gmt numbr E9-J-8W3.
Opinions stated in rbis pa~r do not necessarily represent tie offici~ ~sitinn or poficy of tie U.S.
Depament
of Lahr.
FINAL
Gender
.
REP
Differences
ORT
in the
of Young
Quit
Behavior
Workers
.
Audrey
Light
Unicon
and LfanueEta
Research
Third
Floor
10$01 National
Los Angeles,
hfard
Grant
\vas funded
Number
the oficial
or policy
Opinions
stated
90064
470-4466
30, 1990
by the U.S. Department
B9-J-8-0093.
position
Boulevard
California
(213)
This project
Ureta
Corporation
of Labor,
Bureau
in this document
of the U.S. Department
of Labor.
of Labor
Statistics
do not nemssarily
under
represent
Contents
Executive
iv
S u“mrnary
1
Section
1:
Introduction
Section
2:
The
Section
3:
Characteristics
Statistical
4
Model
of the
Data
.
10
.
Section
4:
Estimates
for the
Section
5:
Estimates
for Selected
Section
6:
Conclusions
Wll
Sample
Sub-Samples
32
43
45
References
.
Appendix’
18
47
.
List of Tables
1
Characteristics
2
Number
Distribution
4
Job Duration
of R-ens
Re=onfor
5
Characteristics
6
Estimates
inthe
Job..,
at F]rst ad
First Yearof
W&bull,
\feans and Standard
Estin]ates
Survey)
Deviations
ad
. .
15
. . . . . . . . . . . . . . .
17
Workers
for
. . . . . .. . . . . . . . . . . . . . . . . .
hfodels
F]rst Jobs Begin or Are in
. . . . . . . . . . . . . . . . . . . . .
Workers
.
Whose
22
Careers Have
in the First Year of the
. . .
23
for ~Tnobserved Heteroge],e.
. . .
First Jobs Begin
. . .
in the Next Six Months,
. .
26
Given Current
. . . . . . . . . . . . . . . . . . . . . . . .
Hazard Model with Comection
20
\vit,h Correction]
Have Not Started or Whose
of Job Separation
.
First Jobs Begin or Are ill
(SAhlPLE:
14
by
Whose
Men Whose
13
. .
. . . . . . . . . . . . . . . . . . . . . . .
Cm~rs
for Modd
for Gompertz
Job ad
Models with Correction
in the First Year of the Survey)
Probability
Current
11
Job Char-
H ~ard
Hazard hiodel witk Correction
Workers Whose
Tenure of XMonths,
. . ..
First Jobs Begil~ or Are in Progress
fo? Gompertz
or Are in Progress
Estimates
Personal
for Leaving
and BOX-COX H~ard
(SAhIPLE
. . . .. . . . . . . . . . .
ity (SAMPLE:
Condition
. . . . . .
. . . . . . . . . . . . . . . . . . .
. .
Women
the Survey)
First Yearofthe
or Whose
by Remon
(SAMPLE:
Progr=s
inthe
. .
Lwt Interview
Heterogeneity
Survey )
.
. . . . . . . . . . . . . . . . . . . . . .
for Unobserved
Sample
,, . .
with Selected
WeibuH, and Box.Cox
for Gompertz.
Not Stinted
11
Lwt
Heterogeneity
Estimates
10
Eqerience,
for Gompertz,
Unobserved
s
by Workers
for Job Separations
and Prior
Leaving
Progress
Lmt hterview
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
4
at First md
of Jobs Held Per Ye=
acteristics
.
of Workers
for Unobserved
27
Heterogene-
6
ity (SAMPLES:
24t031)
12
‘Early-
ad
‘Late”
Birth Cohorts
of Men and Women,
from Ages
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Estimates for Gompertz
ity (SAMPLES:
H~ard
Model with. Correction
Five Entry Cohorts
of Men)
ii
.
.
for Unobserved
. .
..
.
29
Heterogel]e.
. .
34
13
Estimates
for Gompertz
ity (SAMPLES:
14
Estimates
for Gompertz
Estimat=
Model writh Correction
Five Entry Cohorts
ity (SAMPLE:
15
H=ard
of Women)
Hazard Model with Correction
Jobs Ending in a Voluntary
for Gompertz
.’.
for Unobserved
Hetirogene-
. . . . . . . . .
. . . . .
for Unobserved
Heterogene-
Trmsition)
Hazard Model with Correction
.
36
. . . . . . . . . . . . . . .
37
for Unobserved
Heterogene.
ity (SAMPLE:
16
Estimates
JOb-tO-JOb ~asitioni)
for Gompertz
H=ard
. . . . . . . . . . . . i . . . .”.
Lfodel with Correction
for Unobserved
. . . . . . .
39
Heterogene*
ity (SAM”PLE:
Al
Estimates
(S.4NIPLE:
Continuously
for Weibull Huard
Sample
M=ns
of24-31Year
A3
Sample
Sample
Smple
Model with Comection
and Standard ,Deviations
Means
and Standard
Old Men)....
Mess
tions, Job-to-Job
A5
and Standmd
Transitions,
(SAMPLE:
for Unobsemed
Transitions,
md
41
Heterogeneity
or \Vhose First Jobs Begin
. . . . . .“
‘Early”
. . . . . . . . . . .
=d
“Latem Cohorts
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Deviations
(SAMPLE
“Early”
and ‘Late”
Cohorts
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Deviations
for Women
and Continuously
Means and Standard Deviations
Job-to-Job
. . . . . . . . . . . . . . . . . . .
in the First Year of the Survey)
Old\Yomen).
of24-31Year
A4
Workers)
IVorkers \Vhose Careers Have Not Started
or Are in Progress
A2
Employed
(SAMPLE:
Employed
for Men (SAMPLE:
Continuously
Employed
Voluntmy
Workers)
Volunt=y
Workers)
47
Transi-”
. . .
. . .
4i
Tr=sitions,
. . . . . . . . .
.4i
.
.
...
111
.
Executive
Object
.
ives
In this study,
Surveys
*
Summary
we examine
of LAOI
M=ket
about emly-career
1. OverW,
the mhorts of
Experience.
men and women
Our broad
objective
2. \Vhat observable
factors
dence that women,
the most turnover
influence
the turnover
relationship?
of market conditions
3. Do unobservable
these factom
Lon~tudind
the following
questions
or unmeaured
entry
cohorts?
factors
of men ad
account
they are ‘shopping”
Do unemployment
for a significant
for women thm
Do continuously
or job.t~job
In particular,
by such meaures
women chan@ng
than workers who intemupt
Are volunt~Y
of each gender?
is there evifor a durable
of family respmlsibil-
rates and other me~ures
effects on men and women?
more important
behavior
during the early career?
How is’ turnover influenced
have differential
relatively
4. 1s the turnover
turnover
is to tidress
quit their jobs because
as well as men,
ities u marital status and the birth of a child?
m=ket
in the National
mobflity:
which gender undergoes
employment
young
employed
for “men?
amount
Are,,
,
with succwsive
workers
of turnover?”
exhibit
birth cohorts
a different
or labor
pattern
of
their careers?
transitions
caused by a different set of factors
than other tYPes
of job separations?
MetllodoIogy
Although
we begin with descriptive
estimates of discrete time proportional
rate of job
‘Ytilure”
proportional
huard
at a particular
model,
analyses of the samples, most of our conclusions
h=ard
models.
A hward
level of tenure condition
a vector of covariates
is =sumed
model describes
upon survival
to =t
=e based on
the instantaneous
to that level.
multipficatively
In a
on a bwehlle
hward.
After some experimentation,
from a Gornpertz
By estimating
variates.
we settled on the ~sumption
a discrete time model,
we mn readily flow
the effects on the h~ard
a wages, unemployment
rates and mmitd
heterogeneity.
we control
Nthough
status.
can ~timate
mises
for these factors,
the proportion
for the. presence of time-varying
rate of factors
We
dso
correct
or unme~ured
we ensure that our other wtimat~
of men md
women
who ue
careem
had not started
the
job-,
fztors
presence
md
of ~nObserved
mmket-spetific
that sfso influence
~e
‘movers”
CG
that et,olve over time. such
fo~
for a wide mray of’ pemon.,
there me Hkely to be unobserved
By correcting
h~ard
distribution.
That is, we estimate
acteristi~,
that the bm~ne
unbi=ed.
md
da-
turnover.
h addition,
‘stayers”
we
for unobsemed
rewons.
Findings
1. W?hen we look at au workers
in progress
during
the year
have a higher h~ard
and only 46 percent
2. Nlen and women
unaffected
of the first interview
rate than men.
respond
Furthermore,
very differently
hemming
The huard
by their m~itd
to f~ily
mmried,
first jobs
began
or x~,ere
), we find that women
we find that 58 permnt
of the women
re~ons.
d~acteristics.
Among
the fu~ sample
and the b]tih of a ch]ld dl lower tbe h~ad
rate of women f~s
status.
or whose
(the ‘fuU s~ple”
of the men are movers for unobserved
of men, being maried,
job separation.
u.hose
when they get married,
The h~ard
rate incre==
of
but it is otherwise
if they have a newborn
tild,
but not if they have older &ildren.
3. Both men ~d
women appeu
of both genders fd
is spent working,
to eng~e
with incre~ed
incre~ed
prior experience,
tenure, and incre~ed
4. For women, there am lmge differences
The differences
becomes
among
mr insigtifi~nt
in job shopping.
factor
song
incre%es
saple,
in the proportion
the h~ads
of time that
birth cohort md a late birth cohort.
less pronounced.
the late cohort
“
the fti
wages.
between an e=ly
men are considerably
Among
of women,
Unobserved
ad
heterogeneity’
the only important
determinmt
of women’s
of a newborn
turnover that may not be known at the time of hire =e the presence
child and the act of becoming
to be converging
toward
the turnover
5. There me dso importmt
differences
marital status, becoming” married,
important.
For women,
married.
bebavior
w,e find that w,omen appear
of men over time.
among successive
md becoming
the effect of a newborn
are the effects of prior experience
Overd,
labor market entry cohorts.
divorced
For men,
appear to be growing increwingly
child is becoming
incremin~y
important,
=
and wages.
.
6. Among
continuously
ing turnover.
Furthermore,
movers for unobserved
7. M~y
workers, family chuacteristics
only
16 percent
trmsitions.
characteristics
are less important
of the men and 28 percent
in explain-
of the uromen are
reaons.
variables that are important
and job-to-job
family
employed
determinmts
h general,
of job separations
thsse transitions
such ss ed,ycational
attainment,
do not explain voluntary,
are less influenced
becoming
married,
by personal
and
and prior experi-
ence.
Implications
The study
provides
a wealth of dettied
a][d !vomeI], but we wish to highlight
.
unde~one
is more difficult
for employers
a pool of men.
This is because
cannot be observed,
to identify
a luger
not observed
by employers
the pr=ence
of a newborn
child.
erogeneity
by the ftil s~ple
non-quitters
proportion
of women
of young men
mong
the younger
women.
among
are movers for r-ens
that
by two importmt
birth cohorts,
vi
becoming
factors that
mmried
since unobserved
In general, we conclude
women look very much fike men in their turnover behavior.
that it
a pool of women thm
at the time of hir+namely,
is reversed when we focus on separate
is unimportant
behavior
of workers, we conclude
tiong
and because female turnover is influenced
are generally
2. TMs conclusion
the turnover
about
four key implications.
1. When we look at W trmsitimrs
.
information
and
het-
that younger
.
3.” Ftily
r=ponsihfities
tions among women.
(eipeciWy
the birth of a chfid) are not primary
In fact, the h~ard
in the late birth cohort
rate of continuously
are the only ones that incre=e
more Ekely to leave their jobs when they get married
but both of these factors
are short-fived
employed
causes of job ~parawomen md of women
w,ith the birth of a child.
Women
and when they have a newborn
me
find,
by their very nature.
t
4. We find that both
men ad
women
increasin~y
high quality job matches
by investing
in job-specific
exhibit
x
signs of job
shopping.
they gain experience,
That
is, they locate
and they lock into those jobs
.-
skius.
.
.
Section
The days when women automatically
a child are long gone.
I
their caee~,
however,
Nthough
increaing
1:
Introduction
withdrew
it remains
from the labor force upon marrying
common
for women
numbers work continuoudy
that young women
who are not planning
throughout
with young
their addt
career interruptions
chiltien
or ha>.ing
to i=terrupt
lives. There is concern,
are unable to signal their
.
intentions
women
costly
to potential
employers.
have higher average
to women,
expected
At issue is whether
they me quitters.
\vsong to assume
turnover
since trtin’ing,
to workers who ~e
rates than men.
promotions,
‘female”
Such statistid
with “qtitter”
because
discrimination
and even the jobs themsel~,es ae
young women ,me denied valuable opportunities
Not only w,o”ld employers
that young
for a durable
or upon redlzing
may simply equate
would
be
often unavailable
to quit.
]ike]y to quit. their jobs—not
‘shopping”
Employers
men a,e
not quitters.
necessarily
employment
be wrong
~sume
be
men are extre”lcly
from the labor force, but because
They may quit upon discovering
that their current job is not ~ good ~ had been hoped.
prone than men to leave the labor force, the f~t
employers
to take such a \,iew, hut they !i,o”ld
Ear]y in their careers,
to withdraw
rdationship.
because
t]le~ are
a better job
While women are more
is that young workers are Ukely to quit their jobs
regardless of their gender.1
Although
remtin.
,
we can dismiss the notion that women quit ad
Do young women
quit more often than men?
babies, or do they dso engage in joh shopping?
of char= teristics
quittem
thm
Longitudinal
Surveys of Labor
for all job sep=ations,
of whether
‘Of course, young men d= leave the labor f.,..,
about their employme”l
difficult
women
quit primarily
can only observe
for them to identify
To answer these questions,
hlarket Experience
regardless
Do young
Given that employers
at the time of hire, is. it more
the retie non-quitters?
men do not, a number of qu=tions
we u=
(NLS) to estimate
1
non-
data from the National
employed,
primarily m return to school or b{caux
prosp~ts.
a handful
the fem~e
proportional
the worker is subsequently
to have
h=ard
models
unemployed,
they are discouraged
or out of th,e labor
force.z
b
words,
We estimate
separate
of turnover.
We dso
each gender.
This enables us to identify
and tvhether mhort
sub-s~ples:
models
other
w,e view the data from the employer’s
for men and women in order to compare
estimate
separate
models
whether later cohorts
effects are gender-specific.
five cohorts
for an “early”
Flndly,
ad
each gender’s
a slate”
exhibit
we adyze
perspective.
determinants
birth cohort
within
different turno~,er patterns,
the behavior
of four additiond
by their year of entW into the labor force, jobs that were left volunttily,
jobs left to start a new job, ad
sub-sampl~
of continuously
employed
workers.
,
The notion
pmticipation
continue
that women comprise
h= b~n
to withdraw,
rmponsibifiti~,
committed
the labor
number
force
of women
group char~terized
suggests
that even married
to the labor force.
women incre~ed
either
temporarily
work continuously.
women
Atilable
short ad
rate by over 20 percent
me unwi~ng
be able to identify them ~ non-quitters.
to invest in ski~s that will depreciate
=sotiated
with w~e
(SmdeU
and Shapiro,
beneficial
to workers who do not pla
the worker’s
Women
=sume
behef
who plm
to k~p
non-qtitters.
thtir jobs
c=~r
of course,
If they ~e,
Ofek,
relationship
but
incre=ingly
find that 25 to 34 year old
during the 1970s.
to keep their jobs, then it is important
during their absence.
1982; Cox,
pemit
interruptions,
age have grown
opting
1984).
will I=t
will .be denied investment
their careers
Since skiU investments
for relatively
Job-specific
to quit, but such inv=tments
that the employment
they me planning
The question,
ad
women
household
do not dmys
Women who plan.to interrupt
gowt h, these women me essentidy
1978; Mincer
to fulfi~
infrequent
for example,
If women who plan to rem tin in the labor force dso pla
that employers
data
labor force
While may
or permanently
of child bearing
Smith and Ward (1985).
their participation
by sporadic
W]lfis (1977) w well ss others.
such women horn th,ose who experience
us to distin~ish
the evidence
refuted by Heckman ad
from
a l~ge
a homogeneous
=e
flat wage profiles
inv=tments
=e
require that the firm shares
(Becker,
1973
opportunities
H=frimoto,
1981):
if their employers
interruptions.
is whether young women who remtin
then they differ dramatic~y
in the labor force are =tudly
from their mde
counterparts.
A number of
‘Although wc have in mind worker-initiated =parations, we realize that quih, fir=, and layoffs are no} cIearIy
re=ns for leaving
d~tinguishable mnceptudIy or empirically. O“r only rKouw is to m*e u= of workem, reprtd
their jobs.
2
.
studl~
(Bmtel
and Borj=,
1981; Topel
and W7ard, 1985; Topel,
the rapid mobility
that men typicdy
undergo
job match.
workers
tremendous
These
they ‘invest primarily
to exhibit
rderence
dso
enjoy
in sk]lls that
similar job matching
are portable
that women
wage grow,th,
investments
would be nonoptimd
1988)
across jobs
but the evidence
(Light,
1988).
beha.~,ior, then the quitter label cm
are denied trtining
1987,
document
early in their careers as they search for a durable
to their age rather than a discriminatory
accusations
1986; L]ght,
inference
in job-specific
about
suggests
If young women
that
pro~re
be construed
as an accurate
their gend~.
Wtihermore,
skills would be unfounded,
since SUA
“
Efisting
Iiahn,
gender-related
1981; \Vtite md
for both the worker md the employer.
turnover
Berryman,
studies
(Banes
1985; Donohue,
that women who change jobs are indeed attempting
educated
as w,e~ as men.
about
studies pre-date
determinants
1986a, 1986b;
to improve
summarizes
differences
models.
md
1980; Blau and
1986) prm,ide
For example,
wt econometric
and female turnover
and Section 6 conttins
model.
behavior.
our conclusions.
.
3
hints
better
quits are shown to improve
unexplored.
of men’s and w,omen’s quit probabilities
we discuss our econometric
in mde
Lleitzen,
their match.
among women is left l~gely
or ftil to utifize state-of-the
In the next section,
1974; Vlscusi,
These studie~ do not shzre the focus of the proposed
so the issue of job shopping
conclusions
of huard
Jones,
women me found to have higher quit probablfiti~,
wages of women
howmer,
md
r-eirch,
Furthermore,
are suspect
the
many
because
the
techniquw.
Section
3 describes
Sections
the data set ad
4 and 5 present estimates
Section
Most
of the earher
turnover
apprOach that h= be~me
I
tions.
2:
studies
from a~egating
of time-vmying
re~essors.
their hfetime
this context,
the strem
lg$&
Donohue,
benefit ,of continued
mr underlying
The theoretical
the stream
lg86b)
&o
time model)
n.ages &age
spells, but those changes play a key role in determining
unobserved
heterogeneity,
of duration
dependence.
The aggregation
md computations
proportional
The second methodological
of a continuous
=e
h~ards.’
workers’
to
of their skills with jobs.
In
the worker reetiuates
difference
N[odels that do not
is that we include
to do so may yidd
time model to a discrete
the
many times within employment
that we pefiorm
is described
the h~ard
function
The propoitiond
model (in mrrtinuous
of modern
corrections
bl=ed
for
‘=timat~
below.
time model is retiiy
greatly simphfied-when
h’~ards
time
efforts
the spe~’s duration.3
since it is we~ know,n that ftilure
The type of correction
~odel~,
a discrete
dlo~v for the presence of time x,arying regressors ftil to capture the essential ingredients
theories of job matching.
func-
paid to him or her on the current job and
either wage path chang~,
Not only cm’ioth
h==d
to allow for the pr~ence
describes
‘match=”
Ttre use the
~~timate h=ard
First, we intimate
on turnover
for optimal
of w~es
wage offers; whenever
employment.
continuous
logit models.
which is to =timate
respects.
literature
earnings by searching
a worker observe
of outside
1 estimated
lg86a,
differs from theirs in two important
model (obttined
mtitize
cited in Section
Model
standard for the analysis of turnover,
Three recent studies (Meitzen,
but our approach
The Statistical
is restricted
performed—
to the fatily
time) can be expr=sed
of
w
A(t; x) = A.(t) ap(xp)
where f denotes time, Ao(t) is m arbitruy
ad
~ is a column
of discrete
survid
b~efine
vector “of parmeters.
data is appropriate”
that is, when we do not obswe
h~~d
function,
As noted in Kdbfleisch
when the surviti
the exxt
survival
x i3 a row vector of cowiates,
ad
Prentice
time is subject
time (mising
(1980),
a model
to interval
Wouping—
from an underlying
continuous
movement, we capture ~anges
3Slnce we analyze job-to-”o”employ merit trmsitions u WCU= jo~t~job
vd”e of “unemployment by Ietti”g mmital and child-bearing status change tith time.
4
in the
.
distribution)
=e
but rather the interval
k + 1 disjoint
interval
intervals,
Aj in whichitfds, where Aj = [aj-1,
aj), ~ = 1,...,
~+ 1
md QO= O, and Qk+l = m.
In the discrete
time model,
the hazud
at
Aj is defined as
P(T c AjlT > aj_l)
where T represents
the completed
duration
= I-
of a
ew(x, -I 6)
~~,
employment
speU,
~.j=~p(-~,~o(u)d~)
aj-1.
md xj - I is. the value of the vector x at time
In printiple,
l=ge
number
we could atirnate
of parameters.
@ with no reference
One alternative
to the set of incidental
of the =sumption
of proportional
along these Hnes.
First,
that a spe~ w.i~ end.
dependence
potential
presence
Tan, 1985).
constraining
Second,
a simple
h=md.
with parameters
time model dews
that tbe vector
interval to another,
at time aj_l
us to hmdle
x is constmt
the vector x takes on the vdu=
in the sample.
This
function
aris=
for the
second
(W=d
to be estimated
h=ad
tenure
ad
by further
from a Gompatz
C=
6 >’0.
covmiat=
a time interv~
Xj-1,
XSumed
at length in Swtion 4
5
in a convenient
way. By
Aj, but thatitVari=from one
be readily incorporated
Aj. We define the intervals to be thr= months—a
is tiscu+
our estimates
of the lk~hood
time vmying
within
time varying covmiates
‘This choice of h=ard
tenure on the hkehhood
6 and 7:4
Ao(t) = 6eTt,
=suting
because
we ignore the (possible)
across individuals
that the baefine
estimate
made possible
we want to correct
condensation
We usume
.“
The discrete
function
ad
why we chose not to proceed
we chose to reduce the number of paraeters
the b-ehne
distribution
b~ard.
function
AA, (a procedure
the effect of current job
heterogeneity
us from perforating
the likelihood
There are two rewons
a partial likelihood
b~ehne
of unobserved
Therefore,
parameters,
hazmds).
By mtimizing
with @ but it would involve an unusually
is to condense
we want to examine
of the underlying
issue prevents
the set of AA,, together
to rem~n
into the model.
const~t
That is,
Within the interv~
time period in which virtually
dl =pects
of a
job mat&
are Ekely to be constant.5
In duration
models
parametric
distribution,
(Hechm
and Singer,
nonpmametric
mtimum
estimates
1984a,
methods.
hkefihood
the NPMLE
where unobserved
estimates
of the underlying
Heckman
mifing
distribution
is specified
Estimation
of the model
which the number
burdensome.
(1988)
r~ative]y
par~etcrs
of unobserved
to the specified
1983).
limited
individud
used in Lillmd
to have two support
points,
using the NPMLE
estimator
points is determined
One alternative
and Wtite
developed
by Heckma
on observed
individud
the reqtired
distribution
we allow for the presence of unobserved
of proportional
of the b~eltie
shifts in the hazard function
h~ard
is expmded
is, the shift factor is now equal to exp(x@
Pq = 1: h
is required,
involva
Hkehhood
flexibdity
Behrman,
qtite
and Sickels
is a~.tilable,
a
The h’LS prol,ides such
that we expect
that a finite mixture
to avoid bi~~
in the estimat~
of
paaeters.
Consequently,
~,
heterogeneity
and Singer—in
heterogeneity
a wealth of information
compromise.
computationdly
in Taubmm,
of individual
,
where a finite miting
during estimation—is
findings reported
estimates
components.
(1987)
for the finite mixture will do, the job.
factor”
of a nmrpmaetric
heterogeneity
specification
strrrcturd
is to use
MOnte Carlo ~periments-that
parsimonious
points shordd provide
distribution
We befiet,e ttis is a reasonable
that when a very rich set of memures
with two support
to follo~v a known,
rather weU, but it does not yield acceptable
On the other hand, preliminary
indicate
and Mchards,
and find-from
the strate~
of support
is resumed
and Singer ( 1984b) prove the insistency
distribution
We have chosen to fo~ow
Trussell
(NPMLE)
structural
heterogeneity
have been shown to be sensitive
1984b;
estimator
individud
particular,
the estimation
function
is mtimized
by specifying
in the form of a finite mixture.
+ @q), where 6g occurs
sinm.a
only, OZ and
with r~pect
with probablhty
normtization”
P2 = 1 — PI, ~timation
of two parameters
heterogeneity
of the du-
.of the puameters
to a, where
P, = exp(-
That
Pq,such that
of the residuds
of the fitite
PI. To =sist numeric~
a
The “shlft
to include m error term, Og, that takes Q vdu~.
we set g = 1,2, ad
we set 6, = O. Since
indlvidud
mixture
cOn?ergence,
the
~p(a)).
‘Mtho.gh the data permit “s to define intervals ~ short u one month, we found that on+month and thr~
month inkrvds yield virtually identicd ~timat~,
We have optd to “se thr~month
i“tervds b~ause they lower
computation cmti considerably.
6
>
We now describe
the problem
sample”
the fikdihood
of initi~
to individuals
conditions
function
discussed
that emerges
model.
by Heck man and Singer (1984a),
To tinitize
we restrict
the “ffl
w,hose mreers ha~,e not started or whose first jobs begin or me in progress
during the year of the first interview.
at the time of the first interview.,
Because some workers’ first employment
the data set has left truncation
begins md ends before the first survey are =cluded
still in progress
from this hazard
are included).
(i. e., wrorkers whose first. job
from the sample,
The data set is not left censored,
spells are in progress
but those whose first job is
since u,e observe
current tenure
,
for employment
contribution
spells that are in progress
to the hkeUhood
time of the first survey, ad’
function
at the time of the first survey.
of a spe~ that had b=fi
that lasts to tinle t in interval A,,
~P, ‘fi’AA,
q=l
The contribution
,
(
[
in progress
~,,.
P+e, )
)
j>m,
function
for m months
the
at the
is
.xp(x, -, B+Eq) ~ _ ~exP(xr-,
to the hkefibood
Consequently,
of an employment
)
.
spell that starts at the initial
survey date or later, and ends at time /’in interval A, is gi~,en b~
:P, ‘fi’AA,
(
q=l
Findy,
the contribution
exp(x, _,5+o,
) ~ _ ~exP(xr-lo+
(
@q)
~,
)
j=l
to the Hkehhood
of a
employment
survey date or later and is in progress at the time of t be l~t
)
.
spell that begins
at the initial
survey ( i.e. of a censored
spell). say
at time a~, is
M,hen an individ”d
tidition
to the first job,
the second
is included
contribute
in the sample,
W his or her observed
to the Ukehhoqd
or the third c=e just discussed,
depending
function.
employment
The contribution
on whether
5Pe11s, in
corresponds
to
the ending time of the spe~ is
known or the spe~ is censored,
We specify
vidud
within
the unobsewed
individud
heterogeneity
and across spells, but independently
we could ~sume
that unobsemed
heterogeneity
component
distributed
components
i
to be the same for each indi-
across individu~s.
are independent
Alternatively,
across spells for a
~ven
individud
pends
on w,hat is being
time-invariant
constant),
(Tuma,
factors
Han?an,
ad
captured
that make some
then our formulation
spdls
for a given work=
workers
spells, mating
obser~,ed job and, alternatively,
sepmation
mat&
qutity
other band,
heterogeneity
is the source
a significant
is time-invariant
We find the latter.c~e
coefficients.
consists
low h~ards
we estimated
for unobserved
heter~eneity
factors
and “sta.vers’’-i.
heterogeneity
components
hazard models
heterogeneity
me~ured
~d
(by gender)
including
jobs.
If the hazard of job
(see Ffinn ad
Heckman,
1982).
both me~ures
e., the same worke~
on their first jobs.
of future job length have negatis,e
we conclude
that the smple
who have long future jobs
significant
effect .6 Cle=ly,
a short pand introduces
that a longer than average first job must be followed
panel length for men is two, jears
t be duration
coefficient
of first jobs and subsequent
that there is weak evidence
-ong
the men. b
a bi=
~so
job duration
jobs me negatively
(at best) that match qufllty
Therefore,
f=tors
a
and
This may wpltin
is the source of unobserved
fight of the very strong evidence that time-invariant
.,
h=
toward finding
for the men, which suggests
correlated.
have
jobs h=
by a shorter than average second job,
shorter than it is for women.
on subsequent
of Women
For the men, the length of the second job
positive,
why we find a positive
On the .
that make some workers movers and others stayers.
rate of the first job and the average length of dl subsequent
the mtimum
the
~ the length of the second
no effect on the hazud
but marginally
else
on the length of future jobs Jvmdd suggest that the source of
to apply to the u,omen:
of job sepamtion
is
of the length of subsequent jobs, then we can infer that
Since longer future jobs result in a lower hward,
of “movers”
unobserved
as tile average length of all subsequent
coefficient
de-
everything
that dlo%.s for independent
Future job length w=
of unobserved
approach
choice.
heterogeneity,
for the first jobs is independent
of each
If the source of heterogeneity
If, on the other hind,
no corrections
length of future jobs w a covariate.
suitability
components.
then a specifimtion
is a better
The
mo~.ers and others stayers (holding
is appropriate.
To test for the source of unobserved
for first employment
1979).
by the unobserved
reflects the quafity of the job match,
=ross
Groeneveld,
that
we conclude
heterogeneity
are the source of
‘The coefficient on “average length” is 0.019 and lhe norrnd statistic is 1.69; the coefficient on ?eng?h of the
_o”d
jobn k 0.009 md the normal statistic is 0.899. For the wo”en, the mefficien~ (nomd statktic.) m= -0.OX
(-1.80) and -0.041 (-3.34 ), respectively.
8
.
unobserved
heterogeneity
of the unobserved
among
compq~ents
the women,
however,
is justified.
9
we befieve that our choice of specification
Section
We adyze
s21 a~rtildle
3:
Characteristics
of the Data
years of data from the young
The men were fo~owed
Surveys.
men and young
from
women
cohorts
1966 to 1981; although
National
Lmrgitudind
women’s
survey is still in progress, only 1968 to 1985 data me mrdyzed
of the
the young
.
identifying
the date when =ch
pmticipant
entered the labor force.
in this study.
We begfi
To avoid analyzing
by
short-term
,
jobs that ~e
foUow,ed by a return to school,
he or she begmr a period
date occurred
hourly
of labor force pmticipation
in our sample.~
wage is reported.
III addition,
In subsequent
sections,
are in progress
continuously
employed.
Table 1, whl~
compares
men me more succ=sfti
tbe average mm
we present
h=ard
transitions
model
a lower w~e.
counterpart,
trmsitimrs
we summtize
than women in locating
Job seuiority
prior to its inception,
one job for which
~vomu
is equal, so the differenw
reported
1.7 ye=s
of additiond
in ~perience
over a sfightly longer period of time (9.7 versus 9.1 yems),
=
have
of these samples.
suggests
c=eer.
of experience,
is first observed,
than did the ma.
begin or
by workers who
high paying jobs during the ealy
but she h=’ only 0.3 yeas
her current job with more prior experience
which ae
at thtir first and I=t interviews,
the typid
undergone
or whose first jobs
undergone
e~h
he is 20.6 years old, h=
When
for (a) all transitions
(b) only those transitions
workers’ char=teristics
$6.29 per hour (in 1982 dollars).
estimates
have not started
and (d) ti
In this section,
is first interi,iewed,
older than her mde
H that
+
at the time of the first survey,
(c) only job-to-job
18 months.
the worker must hold at le~t
by the workers in these two samples whose c=~n
b=n
that would lwt at le~t
cmeer until
These criteria. yield a sample of 17,361 jobs held by 4,600 men and a
sample of 15,372 jobs held by 4,490 women.
voluntary,
on a worker’s
\vhile the suri,ey \vas in progress or no earlier than six.years
the worker remtins
a
we dld not start the cloti
mems
women
but they receive substatidy
a year
and she earns
that the worn-
On average,
When
and earns
she is almost
experience
that
beg=
we observed
1=s wage
‘Although twmthirds of the partitipa”ts began their careers after the NLS begin, 622 mm and 61? women had
bwn worting for more than one year when they wre fimt intertiew~.
We deleti the .mfll number of workers
who hti b-” worti”g for over six y-r. because it is difficult m determine when they first entered the labor force.
10
Table
1: Characteristics
at First and Lmt Intemiew
MEN
WOMEN
Beginning
of Panel
End of
Beginning
End of
Panel
of PMel
Panel
20.6
29.6
10.8
21.4
2.0
31.0
4.1
0.7
4.86
3.7
6.32
Age
POtentiaJ experience
Tenure
Red hourly wage a
1.7
0.7
6.29
9.77
3.8
li,361
Jobs per worker
Number of jobs
Number of workers
11.7
3.4
15.3i2
4,490
4,600
a In 1982 dollus.
gowth:
men’s wages increwe
only 3.1 percent
the l~t
The=
per year. Furthermore,
numbers
Table
more turnover.
while women
Men are observed
average
and industri~.
average of 0.46 jobs per yeu,
refer to obsemed
Sub-sapl=
jobs
divided
of married
ad
are indistin~ishable
the Mgh school
h= no effect.
s.ob~=rv~.
by
0.4 yems of tenure at
than his female counterpart.
to their early labor
3.4 jobs
holding
in 9.7 yems.
Table 2 shows tlLe Ilunlber of jobs held per yea
SOUPS, occupations,
mtegorier
md less experienced
that men recei$-e a higher return
undergo
paiod,
comparison,
an additiond
incre~e
that while men move into durable, high paying jobs, women either ftil to
>.
matfies or do so much later iu their carars.
1 indicates
nine yea
while women’s
srr~est
good
that they dso
beyond
the average man h=
inter~,iew. despite being a year younger
locate equally
a
by an average annual rate of 6.1 percent,
The first row indicates
force activitiw
and
an average of 3.8 jobs over a
As m
alternative
to this simple
by workers in various demographic
that the sample of 4,600 men holds
while the 4,490 wom,en hold 0.43 jobs per year; these numbers
by the sum of job durations,
white
men average
fa
from the full s-pie.
level appems
Men who are uniotized
fewer jobs
Among
to reduce turnover
dso
so they understate
=d
per yem,
both
men ad
mtuaf
but women
women,
reported or inferrd and =t le=t one wage is reportd. The me-rd= refer to okervd
jok, m they understate turnover m well.
11
in thew
education
Uving in the South or in an SMSA
hold fewer jobs per year while, surprisingly,
job ~e simply job that appear i“ the smpl-i.
turnovers
e., jobs for which starting ~d
uniorilzd
cn~ng dat= uc
of average jobs per year reported in Table 1
r
women hold fm more thsn average.
in construction
and agriculture,
finding is unsurprising,
of short-term
mde
sdf-reported
I=t job.
percent
occurs
in these categories.
This
for at least one job.
differ draaticdly
reveal mother
rrn~pected
sep~aticms
to their f~ily
men.g
characteristics
The NLS repeatedly
prevalent,
cm turnover,
inked participants
we cm ~sociate
for both men ad
When
at similw
re~orrs
working
Women
job separation.
conditions,
consider
why they left their
a response
with roughly
27
of tke women provide
responses.
a
Men
attribute
women
to report
Job dissatisfaction,
and c-workers,
23.7 percent
of thtir
for only four percent of separations
more Ekely thm
the discovery
which encomp=ses
is the most common
reported
re=on
women.
men ad
women
levels of aperience
change jobs
ad
for the same remon,
tenure, ad
w’12 end, in adtitiorr
durations
to why they begin
and memr experience
Since it wi~ l~t
to why they
“Table 4 pr=ents
for jobs dwsified
that the typical mm’s
they are
lead to more durable jobs.
job ended).l”
at the time of trasition
layoff began when be had five years of experience;
whether
a limited number Of jobs according
(i.e., why the l=t
ways. The first row of the top parrd indicat~
‘B=au=
pregnancy is dtern.tively ~ded
a
~e-n
fo,
i=
,OFor ~ job to be Agnd
we wish to 1-
whether the transitions
The panel nature of our data enables us to dwsify
completed
both
with which they ae fired or Itid off, but they
for quitting.
f=
we dso
dlstributicnr of th=e
or health, wfile this re~mr accounts
the reamn for their l~t
wages, houm, location,
the frequency
in the frequency
By the same token, men ae
of a “better job x
resrdt:
turnover.
Table 3 reports
in their reported
leads to a great ded
workers hold fewer jobs than average, yet. managerial
is unusrrdly
women do not differ appreciably
aorrg
of turnover
ntirrre of these industries
the effects of observable
non-coding
mount
me rarely observed
groupings
mde
for job separations.
the greated
of the jobs in each sample; 45 percent of the men and 40 percent
rwpcmse
ad
rewons
women
and technid
effects cm female ad
to mamining
Mthough
dthmrgh
.The occupaticmd
mrd female professional
In addition
the industries,
since the cychc and se~ond
employment.
work hw opposite
Among
me~
in both these
job that will end in a fire or
1.15 yea=,
that is the worker’s
~ YamUy” and “health”, thee categoriu are %gregat.d.
di~lution,
thr~ titeria must be meti a worker must rePOrt
evcntmd
a re-n
for leaving his or her l~t job, the l-t job must be in o“r swplhow”, at l.ut on. hourly we
k reportd, ud no other job intervend
months c= have dapwd betwm” jobs.
12
i.e., its startinS md ending dat= ~e
between the tw~and
no mOre than 18
Table 2: Number
of Jobs Held Per Yea by lVorkers with Selected
PersOnd and Job Characteristics
MEN
lobs Per
Ye=a
Fdl Smple
M=ried
Konwhlte
< 12 years of school
> 12 years of school
five in South
five in SMSA
Wages set by union
Industry
Agricdture,
Mining
for=try,
fishing
Construction
Manufacturing
Transp., communication,
utihties
Trade
Finance, insurance, red estate
Services
Pubfic administration
Occupation
Clerical
Sales
Prof=siond,
technical
Managerial
Craftsmen, foremen
Operatives, laborers, service
“Zk/XDJ,, where k = 1 if job j is in ate
bZD,,fZD,,where D,~ i. duration if job
WOMEN
Percent of
Time Spent
in Categoryb
Percent of
Jobs Per
Yeara
Time Spent
in Categoryb
0.43
0.45
0.44
100
0.46
0.38
0.44
0.43
0.5i
66
34
40
0.46
0.40
0.53
100
0.49
0.42
61
3?
0.49
0.45
0.36
40
73
14
O.iO
0.39
0.i5
0.38
0.34
0.57
0.42
0.50
0.28
2.5
1.4
9.0
35
8.7
16
3.7
17
6.4
0.61”
1.8”
0.41
0.27
0.64
0.34
,0.43
0.29
20
5.5
15
8.6
42
6.5
0.42
0.45
9.1
5.4
17
7.5
17
43,
0.39
0.71
5?
25
0.34
0.38
0.46
‘
0.55
Y k, Ootherwise.
s in cate”goryk, Ootherwk.
“ Includes mining and construction.
13
0.34
0.44
0.41
0.54
56
30
73
14
40
3.8
19
2.5
1.0
31
Table 3: Distribution of &wons
for Job Separations
Reported
Rewon
Fired
Ltid off
Quit due to
Hedth/f-ily
.kccordlng
WOL4EN
2.8
22.6
4.3
29.9
24.4
3.2
Job d!ssatisfution
Found better job
School
Other
Number of .iobs
Number of ;,orkers
tenure when he is discharged.
MEN
,3.2
26.1
23.7
26.87.6
3.5
12.9
8.9
4,836
2,049
3,930
1.813
to the first rowofthe
bottom
panel, men \\,hohave been
fired or ltid off acquire new, jobs that lwt 0.96 years, on average.
A number
of interesting
Inore tenure thm
subsequent
longer
they are abetter”
are much longer thm
while women
which, for men, is Ekdy
age, whale pre~mcy
The turnover
non-employment,
behavior
SkiRs, whllethe
First, women
thus far includes
spells that mnnot
of men ad
women,
S~Ond, jObstha~are
istruefor
0f.5.4yea=
Of murse,
factor
th~
we me~ured
su~~w
that
entered intObecause
although
jobchanges
caused by
Ofexperience
the Ukefihood
ftily
women tend
wh~they
of poor hedth—
obligations-incraa
with
are hkely to occur emly in tbe mreers of women.
transitions
be identified.
fmm jobs
we wish to focus On tho~
transitions
unemployment,
is to compare
the job
whI& =e job-to-job.
to identify a job-t-job
the length Of time that elapsed between
14
to new jobs,
Since one of our gods
Given the spotty nature of the timelines, it is not always e~y
For each job,
and their
latter suggests that they firrd relativdy
anaverage
have only four.
to be a more importmt
described
The former comparison
Just theoppmite
Menhave
and family obligations
ad
disch~ge.
tend to have stightly
versus 1.15 years),
average for both men and women.
later intheircmers.
and f=ilyobfigations.
make such traditions,
shopping
injob.specific
after minvOluntary
tr~sitiOns
hdthproblems
Table4.
(1.13’ years versus 0.96 j~ars).
less intensively
better (i.e., 10nger)jObs
tOmdesu&
emergefmm
men whel) they are fired Or ltid off (1.31 yeas
jobs l~t
wOmeninvwt
contrasts
its termination
transition.
date md
the
,.
Table 4: Job Duration and Prior Experience, by Re~On for
Lea\,ing Current Job ad by Re=on for Leaving L-t Job
\\TOhfER
MEN
{umber
Reason for Leavin~
Cument Job
Fired /laid Of
Quit due to
Health/family
Job dissatisfaction
Found better job
School
Other
AU quits
Number
Mean
Nlean
Prior Exp.
1,496
1.15
5.00
of Jobs
1,161
Duration
1.31
Prior Exp.
4.63
201
1.67
1.44
1.?2
1.06
1.61
1.55
5.39
4.27
1,018
1,186
4.42
2.73
5.21
4.43
351
159
551
3,418
1,403
1,138
185
476
3,403
Retion for Leaving
L=t Job
Fired/ltid
hleu
Duration
htem
of Jobs
off
Quit due to
Health/family
Job dissatisfaction
Found better job
School
Other
AU quits
1.42
1.31
1.96
1.31
1.82
1.51
3.95 ,,
4.13
4.66
2.96
4.20
6.80
Xumber
b{ean
Mem
Number
~ Mean
Mean
of Jobs
1,418
Duration
0.96
Prior Exp.
6.85
of Jobs
1,000
Duration
1.13
Prior Exp.
6.58
1.30
1.41
1.82
0.91
1.28
1.50
7.80
6.48,
6.98
4.62
7.68
4.09
931
1,054
299
138
508
2,930
1.05
1.28
3.76
1.13
2.06
5.96
6.13
7.32
4.89
6.95
6.28
207
1,444
1,179
157
431
3,265
,
,,
15
1.59
stinting
date of the next observed job.
de~,iation of 14.47; for wmen,
that are fo~owed
smple.
Fm men, the mem
the mean is 15.70 months and the stmdard
by a gap may in fact be immediately
However,
in an effort
to select jobs
require that the gap be less thm
This selection
criterion
“gap” is 11.07 months,
succeeded
that ue
tiown
de~riation is 23.34. Jobs
by a new job that is not in OU1
to be fo~owfed by another
job,
we
d
three months.
yields a sample of 2,974 jobs for the men and 1,630 jobs for the women
that end with job-to-job
transitions.
The
average
rn~’s
job
the average w,oman’s ends after 1.03 years. In our subsequent
the factors
w,ith a stadmd
ends
when
andytis,
that cause these jobs to end differ born the f~tors
years Old ad
it is 0.87
we will detedne
whether
that cause the full smple
of jobs
to end.
The finrd sub-sample
TO constmct
these s~ples,
since the l=t
Agtin,
interview)
consists
we used information
that respondents
of “workers who me ‘continuously
on the number of weks
spe;t
working,
unemployed,
ye are, not able tO accorrnt for d] the weeks in each worker’s
invmiably
order
that u,e exmine
lose sight of a large number
to SKIN this problem,
the percentage
percent
of time that is known
of the entire pael
to be spent working.
media
we focus
to be spent
is accounted
wOrK]ng is considered
Or non-employment,
workers.
numbem
It revds
successful
men at trmsiting
the sae
tenure.
employed.
However,
In particular,
are conducted
be accounted
81.1 percent
bimnudly.
for ~d
h
‘calculate
an average
of 85.6
of this time is known
are 68.5 ad
percent—s
we
62.1.
We use the
the cutoff point in de fiting
Gaps in the data, rather than unemployment
contined
in Table 1 for the saples
that, among continuously
~onnt
career.
force.
for most of the time spent not working.
ployed workers.
have rOu@y
or out of the labor
(or
That is, anyone who spends more than 88 percent of thek time
Table 5 drrpficates the information
tbm
during the Imt ye=
For the men,
for the women
distribution-89
to be continuously
account
working.
for, and mr avemge
The corresponding
employed
when interviews
on those weeks that wn
of the men’s time-spent-working
continuously
tidltiond
of w~ks
employed.”
into’ durable jobs.
employed
employed
16
em-
workers, women appear to be more
At the end of the panel, men md
of labor market =perience,
cmrtinuously
of continuously
but women have almost
women appea
women
15 months
of
to be getting lower returns to
Table 5: Characteristics
(SAhlPLE
at First and L~t
Continuously
Employed
NfEN
Age
Potential
Tenure
experience
Red hourly wage a
Jobs per worker
Number of jobs
Number
of workers
Intemiew
Workers)
WOMEN
Beginnillg
of Panel
End of
Panel
Beginning
of Pmel
End of
Pael
21.1
2.0
30.2
11.1
21.6
1.6
31.1
11.2
1.0
5.3
10.23
1.0
5.52
6.5
7,79
6.70
3.2
7,433
2.5
2,812
2,343
1,142
“ In 198? doll.rs.
tenure than their male counterputs:
0.76, whi&
is exactly
the ratio of female to mde
the same as the ratio for the full sample.
wages at the end of the panel is
Section
Prior to adopting
the b~efine
hward
4:
Estimates
a unique specification
v,=
alternatively
for the Wll
for the bm&lne
specified
Sample
h~md,
we estimated
to be Gompertz,
Weibrdl,
models in w.hid
and BOX-COX.
The
.-
bxeke
h~ard
specification,
under
a GOmpertz
specification
w=
described
in Section
2.
Under
a Weibull
it is
,
Ao(t) = pk(pt)~-1
with p, k > 0. The b~ehne
huard
u,nder a BOX. COX specificatimr
~vith A2 > Al z 0. .The Box- Cox hazard
md the WeibuH hmards
the above wpre~ion
~ special cases.
yield$ a Weibull
ii defined ~
is quite general in that it conttins
botb
the Gompertz
It is readily verified that setting Al = O and T2 = O in
hazard,
while setting A1 = 1 md
72
=
O yields
a
GOmpertz,
h=ard.
The ~tra
function
h=ad
solution
flexibility
for our discrete
= a function
is dso
avoided
if numerical
pmameters
required
of magnitude,
in progess
we. foUow
to compute
the stand=d
(continuous)
of a BOX-COX hazard,
b=&lne
numerical
the ~adient
derivati~,es me used—a
appro=h
h==d.
To write
of expressing
is required.
vector at each iteration.
practice
we avoided
the Ukefihood
the discrete
Since there is no =~ytid
integration
tests, we found that numerical integmtion
so we obtdned
in the first yea
of initial conditions
smples”
model
comes at a price.
because
Numerical
Of course
inte-
this can be
of the large number
of
md sample sizes involved,
In prehtinary
yields
time
of the integrated
for the integral
~ation
afforded by the BOX-COX h~~d
ftirly
observations
rrf the survey.
and fikely minimizes
large samples
described
estimates
mryway.
earher ~
estimation
time by an order
for’ a sample of workers ~hOse first jObs began
This selection
the problem
For botb
conttined
irrcre~ed
which efiminat-
of time inbomogeneity
men md
in the restricted
for 6,191 workers, and the fi~res
criterion,
women,
smpl-:
about
the prObIem
of the mvirontient,
35 percent
of the ‘full
the sample of men ha
for women me 27,025 observatimrs
18
or were
31,948
for 5,313 workers.
We report
our estimates
four parameters
ad
Gompertz
for the three models
to capture
h=mds
the tenure dependency
have only one parameter
model yields the largest ldue
estimates
for women
the GOmpert~
~timates
Indeed,
of the hketihood
the -timates
BOX-COX specification
6 and 7. “Tile BOX-COX h=md
of the hazard
sch.
function
worse fit. Comparing
Agtin,
“movers”
for unobserved
though
and, therefore,
is the estimate
the estimate
model,
compared
with 74 percent
in the GOmpertz
appem
to be unaffected
by the choice of bm~ne
md
given the high cat
=timates),
specify
a GOmpertz
We estimated
h=~d
of obtaining
for the remaining
a WeibuU model
considerably
better.
TO avoid
h=ard
(generWy
during the ye=
in November)
criteria field
ad,
bawd
(ignoring
are reported
model
we chose to
haz=d
to determine
specifications
whether
are sample specific.
the GOmpertz
model
fits the data
in Table Al in the AppendIx.
mentioned
of the first intemiew.
in Section
2,552 women.
19
2, we begin
by =timating
had not started or whose first jobs began or
The first interview for the men wu in 1966
for the women, it w= in Jmuq
samples of 2,594 men ad
parameter
the WtibuH
for the BOX-COX model,
sampl=:
for samples of workers whose car=rs
were in progres
at 86 percent in the BOX-COX
for the full samples of men and women
of initial conditions
in the intercept
analysis.
for the restricted
The estimates
the problem
models
found
of men who are
Give’n that tbe remtining
estimates
our findings on the relative merits of alternati~,e bwetine
We find the same pattern
the BOX-COX
and the estimates
of 02 (the difference
md “moversn ). The fraction of stayers is estimated
model.
estimate,
of the fraction
term for ‘stayers”
=timates
in this c~e
equal to -10011.8 versus a value of -10045.8 from the
The one exception
rewons
the parmeter
benefit to be gained from using the
we find no sign reversals in the parameter
are very close in magnitude.
th~
analysis.
model yields a value of the UkeEhood function
model.
better
The
and from the BOX-COX models we find no sign reversals.
The results for the sample of men parallel those for women,
Gompertz
that the BOX-COX
fits the data only marginally
are so close that there is no apparent
“in subsequent
h=
the WeibuIl
for the samples of men and women.
results in a considerably
from the Gompertz
of job separation;
Thus, it is unsuwrising
S11OV,that the BOX- COX hward
the WeibuU hmmd
that obttin
in Tabl=
or Februa~,
1968. These selection
..
(SAMPLE
Table 6: Estimates for GOmpertz, Weibull, and BOX-COX H=ard Models
Correction for UnObser\.ed Heterogeneity
Women Whose First Jobs Begin Or Are “in Progrws in the Rrst Yem of the Surve},)
-,
Variable
NOK!VHITE
MARRIED
WEDS
DWORCES
BIRTH
NEWBORN
PRESCHOOL
GOMPERTZ
Normal
Coefficient
Statistic
-.093
-1.68
3.09
..182
1.50
.138
-1.08
-.125
3.74
.243
.287
3.89
4.50
.238
SCHOOL
PRIOREXP
.007
-.206
.41
-9.il
TIMEWORKING
INVOLUNTARY
WAGE
‘P.4RTTIME
-.051
.388
-,96
6:81
-.511
-9.42
i.06
5.i4
UNION
CONSTRUCTION
TRANSPORTATION
TRADE
PUBADhiIN
SMS.4
SOUTH
UNEMPLOYMENT
Y6869
Y7071
Y7273
Y7475
Y7677
Y7879
Y8081
Y8283
7
log(k)
71
Y2
10g(A, )
log(A~)
10g(6)”
a
o~
L% Hkelihood
. Forthe Weib”ll model, t}
_del.
.334
.341
.059
-.383
.426
-.481
.013
.059
.001
-.453
-.106
-.051
-.614
.278
-.123
.607
1.246
-.021
—
.38
-5.32
i.36
-5.94
.22
.1.15
.11
-4.57
-.76
-.33
-3.03
1.20
-.43
i.89
3.31
-12.4i
WEIBULL
Coefficient
-.084
.169
.144
-:114
.252
.288
.225
Nomd
Statistic
-1.30
2.86
1.47
-.91
3.66
2.85
3.93
-4.05
.008
-.202
.49
-8.54
-2.02
6.24
-.089
.3i0
-.545
.320
.309
.051
-9.94
5.34
4.82
.29
-.39i
.396
-3.26
6.60
-4.10
.44
.52
1.13
-6.00
-4.94
-5.89
-9.77
-6.35
-9.17
-6.76
-5.45
—
-.500
.325
.315
.067
-.376
.409
-.473
.023
.053
.002
-.437
-.098
-.033
-.598
.290
-.114
.605
1.220
—
—
Coefficient
-.065
.124
.122
-.082
.267
.319
.204
.004
-.051
-.185
.409
-.524
.026
.029
.014
-.700
-.672
-.871
-1.803
-1.209
-1.912
-1.562
-1.378
—
-.182
—
—
—
—
—
—
—
—
-2.220
-9.00
-1.948
-3.64
22.02
-.182
1.316
k Iog(l
-8529.557
t k the~timateof~
—
-.286
1.193
-8490.83i
is the -timateof
BOX-COX
Normal
Statistic
-.9i
2.16
—
—
1.2$
-.54
3.76
2.96
3.44
.29
-7.18
—
—
—
—
-8.44
-1.55
16.59
-.114
-.009
-17.315
.150
-2.167
-.219
1.173
i
-.99
5.01
-9.09
6.05
4.51
.46
-3.41
7.5i
-4,61
.38.
.88
.26
-4.41
-.72
-.20
-2.90
1.13
-.38
1.i6 .
3.06
—
—
-3.25
-1.86
-.27
1.41
-.8.75
-1.65
14.92
-8486.012
taut term in the Box-ax
.
Table 8 summmizes
the job match, ad
marit~
the variables
the efivironment.
terval.
dummy
To control
me=ure
for women,
past job shopping
the fraction
of potential
and a dummy
&aracteristics
indicating
proved
characteristics
are dummy
nriables
activity
prior experience
\,ariable indicating
status,
part-time
public administration.
estimates
The first tfing
have expected.
married
is not statistidly
but, surprisin~y,
sepaate
12 We ~ci~”nt
variables indicating
Table 9 reports
BecOting
a measure
during the in-
whether
of years of completed
that is wmunted
schOOling.
for by obserued jobs
(TIMEWORK-
the lmt job w=
left involuntarily.
,.
TO memure
lowers the h=ard
hazard model
while incremed
more pronounced
with monthly
with corrections
bce
unemployment
rates
education
do not effect the hwards
effect for women.
lowem the h=md
Nevertheless,
child, although
attinment
hetw one
wmuen, but the effect of a divorce
The birth of a &ld
is not an important
for unobserved
rate for men but h= no effect for women.
rate for men ad
for either gender.
an insignificmt
no effect on their h=ard.
~,ariables
calendar years, residence in an SMSA, and residence in the South.
Being married lowers the h=ard
it h=
~’e
(PRIOREXP),
for market chmacteristics
for a Gompertz
significant
child
prior experience
to note is that personal charmteristics
(WEDS)
a child is
and industry of employment.
Since most industries
.
we include ordy construction,
transportation,
trtie~ and
from thtir jobs if they have a newborn
or women,
indicating
of turnover,
of whether she has a newborn
years of potential
whether
determinants
race and
status,
to have no effect on the hazard,
erogeneity.
a dummy
we dso include indicators
by including
indicating
the worker marries or divorces
we include
we include
of the individud,
of the current job match, we include the (log of the) hourly wage, dummy
union
dummy
whether
for the effects of children,
children. 11 In ~~tiOn,
or preschool
fight
In the first category
\rariabIw indicating
born during the intend;
ad
which include
status; since changes in marital status are fikely to be important
we dso include
ING),
we consider,
predictor
women are more fikdy
the pr=ence
of turnover
lowers both h~mds,
rate for men
of a preschooler
behavior
although
to
h=
for either men
the effect is far
for men.
I, The ~“mbe, of chfidren ag, 18 0, “nd,r ~roved to have no effect on the huard for either gender. Sivce Mde
mDondents were not -kd their cbildre”’s az-,
we cannot include the NEWBORN and PRESCHOOL dummies
-.
in $htir b=ards.
XZPO,the wo~,”, the ~onstr”ction d“mmY dso i“cl.d~ agriculture nd mining. We formed th~ .OmPO.ite
industry becanse only a handful oi women appear in emh category.
21
-
—
Table 7: Estimates for Gompertz,
Weibull, and BOX-COX Huard
Correction for Unobserved Heterogeneity
(SAMPLE
Men Whose
first “Jobs Begin 0. Are in Progress
Vmiable
Coefficient
in the Pirst Yem of the Survey)
~,EIBuLL
GOLfPERTZ
Normal
Modds
BOX-COX
Statistic
Coefficient
Statistic
Coefficient
.031
-.139
.55
-2.46”
.036
-.142
-.295
.66
-2.76
.032
-.237
-3.79
DIVORCES
BIRTH
.133
-.277
.76
-6.12
-.337
.115
-.248
SCHOOL
PRIOREXP
TIMEWORKING
INVOLUNTARY
WAGE
PARTTIME
-.023
-.206
-.326
.423
-.124
-2.2i
-8.66
-4.53
7.38
-2.64
.17:56.
-.035
-.014
-.551
.333
-.203
.693
-4.17
10.25
UNION
CONSTRUCTION
TRANSPORTATION
-.448
.419
-7.61
i.~1
-.477
.443
-6.50
6.52
-.399
.066
-.271
-.293
.017
-.260
.30
-235
-.377
.047
-.275
SbfSA
SOUTH
UNEMPLOYMENT
Y6869
Y7071
Y72?3
Y7475
Y7677
Y7879
Y80S1
-.099
-.055
-.035
.748
1.396
1.004
1.417
2.124
2.387
2.502
-.030
—
—
-9.96
1.21
-3.05
-2.18
-1.10
-3.36
10.85
13.17
-2.86
TRADE
PUBADMIN
-.074
-.084
-.037
.517
.766
.056
.113
.373
.229
-.030
—
-1.41
-1.63
-2.66
7.25
7.17
.46
.71
2.18
i.33
-.13
—
-.092
-.064
-.035
.795
1.412
1.076
1.500
2.162
2.381
2.479
—
-3.20
9.63
9.45
4.93
5.51
6.46
6.44
5.65
—
-.033
—
—
—
—
-13.12
—
—
—
:g(k)
71
—
72
—
10g(A, )
—
10g(A2 )
10g(6)-2.108
a
-1.209
e=
.870
Log Uehhood
-10045.896
. FOrthe Weib”llmodel, th is the estimate of
6.i4
7.54
8.99
8.85
i.85
.14.55
—
—
—
—
—
.14.87
-2.68
12.77
k l.g(p:
22
-4.68
-3.52
-.294
-3.21
.63
-3.74
.143
-.262
-3.88
-1.08
-6.43
4.96
-.017
-.199
.34
-3.92
-1.11
-7.01
-4.73
-.397
.394
-.140
.658
-.428
.419
6.45
-2.10
7.77
-6.40
5.42
-3.44
.80
-2.55
-1.64
-1.26
-.52
-.023
-3.84
-.096
-16.866
-.02
-4.95
-.387
-9.45
-1.781
-1.997
-7.93
1.26
-1.903
.186
-.82
1.209
1.89
8.79
.819
-10079.701
-10011.867
it & the estimate of the constant term in the BOX-GX
—
—
—
“=
Statistic
NONWHITE
MARRIED
WEDS
.6i8
.55
““”Nomd
Normal
..
Table 8: Sample Means and Standwd Deviations
(SAhfPLE:
Workers Whose Careers Ha\,e Not Started Or WhOse First
Jobs Begin or Are in Progress in the First Yem of the Survey)
MEN
Definition
Variable
1 if nonwtite
NONWHITE
1 if mmied
MARRIED
WEDS
1 if mamies during interval”
DIVORCES
BIRTH
NEWBORN
PRESCHOOL
SCHOOL
AGE
PRIOREXP
TIMEWORKING
INVOLUNTARY
TEKURE
PARTTIhlE
~rAGE
UNION
CONSTRUCTION
TRANSPORTATION
TRADE
PUBADMIN
UNEkfPLOYMENT
SOUTH
SMSA
Y6869
Y?071
Y7273
Y7475
Y7677
Y7879
Y8081
Y8283
Number
Number
1 if divorces during interval”
1 if Aild is born during intervda
1 if &ild age one or under
1 if child age six or under
Years of schooling
Years of potential prior experience
Ratio of actual to potential experience
1 if left I=t job involuntarily
Years
1 if works
of tenure
less than 35 hO”rs/week
Log of real hourly wage6
1 if wages set by union
1 if industry is
Construction
Transp., communication,
utilities
Trtie
Pubfic administration
Unemployment
rated
1 if fiving in the South
1 if~ving in an SMSA
1 if year is
1968-69
1970-il
1972-73
1974-75
1976-77
19i8. 79
1980-81
1982-83
of individuals
of obser}.ations
Niean
WOMEN
Std.
De\,.
0.24i 0.431
0.589
0.059
0.492
0.235
0.018
0.150
—
—
0.134
0.357
—
b In 1982 doUars,
‘ Includs mining md agric”lt”re.
‘ Unemployment rate d“ri”g first month of the i“tervd.
23
Std.
Dev.
0.274
0.593
0.059
‘0.446
0.491
0.235
0.026
0.120
0.159
0.325
0.1s8
0.414
2.347
12.989
2.821
0.037
0.220
13.034
25.53i
3.644
0.488
0.094
2.542
0,061
2.015
0.175
4.611
26.694
3.604
0.348
4.031
0.415
0.292
3.010
0.240
0.500
0.380
0.085
2.651
0.161
1.702
0.195
0.101
0.072
0.301
0.258
0.019”
0.052
0.137
0.222
0.167
0.058
0.373
0.234
0.152
0.060
0.359
0.237
9.213
0.417
0.730
2.951
0.493
0.444
11.247
0.377
0.747
3.399
0.485
0.435
0.118
0.323
0.061
0.239’
0.166
0.130
0.143
0.141
0.131
0.118
0.000
2594
50,16i
0.372
0.336
0.350
0.348
0,337
0.322
0.016
0.111
0.127
0.134
0.136
0.127
0.125
0.174
0.314
0.333
0.341
—
a Since we O“IY know that a change occur. betw~n s“cce=ive interviews, the vmi=ble ~“~s
idling between the two interview date.
six-month i“tervd
Mean
2552
48,570
one for every
4.940
4.182
0.337
0.279
3.02G
0.367
0.476
0.396
0.343
0.333
0.330
0.379
The effects of prior experience,
job shopping
md
although,
tenure, and wag=
x we saw in Section
TIME!I’ORKIN”G
=pecidy
workers lock into good matches,
m increase
d-pened
coefficients
by each gender.
About
value).
have a negative
in tenure for both
This is consistent
is considered
to be a good
but especially
for women.
more iucc=sful
in mde
and the industry
ad
dummies
female turnover
in service professions
from their union status.
The coefficients
comparable
for men and women.
in the mnstmction
titinistration.
suggest
for both
&ect
for women.
in a SMSA
dummies
The h=md
h=ard
significant
to the “omitted”
reverd a pronounced
for women.
from lager
-timat-
numbers
period
secula
The smple
rates of men and incre=ing
a figh
curr~t
wage
that some of the observed,
to the types of jobs favored
rates, on the other had,
effect on the h=ard
for both tien
have the s-e
W= characterized
rates for women.
Mthough
commitment
ad
pubfic
for mmr, but h= “no
yems, with the yeas
for women,
for men and a secul=
by decfining
incre=ed
prior to 1968
The coefficients
labor
on the y-
decre=e
in the
force participation
pwticipation
rates cm
rtidt
change in behavior,
to the labor force may have dso
24
rmd me of
is higher for work-s
lower the h=ard
of women in the labor force with no xcompanying
suggest that an increwed
si~s
rate for men.
rmd women.
in tbe h~md
uromen w,ho
md do not gtin job stabi~ty
and is lower for workers in transportation
incre~e
period
refer to rrdon
.4pparently,
of job separation
The fist of regressors includes dummy variables for cdend~
dummies
that
The effect for men may be
and in the south lowers the h=md
Higher unemployment
but do not have a statistically
corresponding
but
of match qudity—
men and women
( e.g., teaching)
mr the industry
and trade industries,
R-idence
genders,
mr
sWIS. We dso find that
in parlaying
is attributable
19 perc~crt of the observatimrs
are crmcentrated
magnitudes
effect
with the notion
indication
but UNIOhT lo~vers the hazard for men and rtises it for w,omen.
are unionized
.
dechnes
P~OREXP
wage offer.
On UNION
difference
experience)
by investing in match-specific
by the fact that they are ,rLlatively
unmndition~
jobs,
presumably
rate for both genders,
into an even higher outside
The
rate dso
in”“absolute
in the current wrage—whiti
lowers the h~ud
to potential
The h=ard
for men (i. e., y is grmter
women engage in
3, men appear to do so more vigorously.
(the m.tio of observed
the hazard rate for both genders.
suggest that both men ad
our
been a fmtor.
.
TO summarize,
the ~timates
differ, but not drmaticdly.
of mobifity
increased
k
for both genders.
can use to predict
turnover
by an important
of a newborn
child.
suggest
that the early
particular,
Although
behavior
job shopping
turnover
of men and women
appears to be an important
we have identified
a number
for each gender, we have dso
factor that is often unknon,n
For this reason done,
behavior
determinant
of ~,ariables that employers
shown that female turnover
at the time of hire—namely,
it may be more difficult to scr=n
is
the presence
for non-quitters
among
women than among men.
Furthermore,
reveals,
we must take into account
it plays an important
role in explaining
of Oz is 1.02 for men and 1.22 for women.
incre=es
the h=ard
the effect of unobserved
turnover,
These numbers
esp’ecidly
imply
heterogeneity.
for women.
The ~timate
that unobserved
heterogeneity
rate by a factor of 2.8 for men (el 02 ) and by a factor of 3.4 for women.
mn dso identify
what proportion
the observable.
The value for a is -0.49 for tbe men and -0.13 for the women,
46 percent
of the sample changes jobs’for
of the men and 58 percent of tbe women ~e
=timates
indicate
However,
factors
that there is a tremendous
that employers
for men. This conclusion
TO detertine
men ad
by the results we obttined
of job separation
who is white and mmried,
The pmbabihties
for the period
13p,
=
.zp(_e=
were evaluated
retions.13
heterogeneity
more important
rates, P~OREXP,
of job separation
1968-69.
p(=))
i.
ad
for women
For the remtining
and female stayers.
the
~.Tce”t
that
ue
for men and women)
for women tha
the BOX-COX
six months (condition
These condition~
‘stayers,m
SMSA.
in Table 10 reports
the probabfities
and Ps = 1 —PI percent are “movers.”
25
a modd
at the mean values of wages,
The top pael
we mrly report
On
probabilities
and fives in =
and for men and the ratio of the former
periods,
These
stayers differ in their quit behavior,
in the following
TIMEWORKING.
by
within genders.
when we estimated
hm twe~~,e years of schoofing,
(separately
We
which implies that
for the periods 1968-69, 1976.77, and 1980-81, for what roughly constitutes
sommne
pmbabitities
that are not captured
for unobsert)ed
amount of unobserved
vmious levels of cnrrent tenure) for both mde
unemployment
‘movers”
women who =e deemed
we comp”ute the imphed probabilities
worker:
re~ons
cann~t control for are relatively
is reinforced
whether
were computed
As Table 9
the
to the latter,
for men ad
Table 9: Estimitei
for Gompertz
H~ard
Model
Correction for Unobserved Heterogeneity
(SAMPLE:
Workers Whose Careers Have Not Stated or W7h0se First
Jobs Begin or Are in PrWr:ss in tke First Ye= of the Survey)
WOMEN
MEN
Vaiable
NONWHITE
,NomA
Sbtistic
CO~cient
Coefficient
Normal
Statistic
0.014
0.42
-0.051
-1.58
MARRIED
WEDS
-0.205
-0.372
0.037
-0.147
1.2i
-2.83
DIVORCES
BIRTH
NEWBORN
0.106
-0.334
—
—
-6.86
-7.20
1.08
-8.39
—
-0.121
0.046
0.379
-1.52
1.19
6.43
0.047
-0.026
1.42
-3.95
-16.02
-7.11
PRESCHOOL
SCHOOL
PRIOREXP
TIME\\ ORKING
INVOLUNTARY
WAGE
PARTTIME
UNION
CONSTRUCTION
TRANSPORTATION
TRADE
PUBADMIN
SNfSA
SOUTH
UNEMPLOYMENT
Y6869
Y7071
Y7273
Y7475
Y7677
Y7879
Y8081
Y8283
Lg(6)
a
e~
Log fike~hood
-0.057
.,0.147’
-0.364
0.272
-0.230
0.534
-0.591
0.398”
-0.300
0.128
-0.361
-0.105
-10.39
-20.33
-8.34
i.19
-8,00
14.58
-14.42
:..10.72
-5.23
4.12
-5.39
..-:3.45
-0.099
0.002
1,067
1.441
0.672
0.961
1.483
1.675
1.780
—
-0.026
-2.293
-0.492
1.019
-30275.431
26
-3.39
0.21
20.27
22.14
9.59
10.99
16.07
17.79
15.06
—
-33.99
-2322
“-4.47
28.31
-0.105
-0.329
0.442
-0.58i
11.03
-19.37
0.407
0.478
0.398
-0.575
0.334
13.10
14.84
4.76
-7.74
10.58
-0.367
0.047
-5.58
1.58
1.34
.3.49
-3.62
-1.81
-4.83
-10.94
-5.30
-10.86
-5.06
-1.52
-1 i.23
0.038
-0.021
-0,375
-0.190
-0.509
-1.2=
0.617
-1.313
-0.6.61
-0.220
-0.014
-1.538
-0.131
1.218
-10.81
-1.63
23.82
-271i6.867
Table 10: Confitimrd
Probability
Given Current
By Gender,
—
Current
tenure
.107
6
12
24
.099
.091
.078
36
48
.067
.057
.049
60
FO
1968-69
blen
WOmen/.Men
,VOmen
o
Of JOb Serrarationin
Tenure Of XMOnths,
jelected
.621
.249
.217
.190
.143
.072
.107
.054
.040
1.067
1.146
.129
.097
—By Gene
Current
tenure
o
G
12
24
.320
.281
.254
.267
.382
.433
.282
.493
.246
.188
.142
.080
.564
.106
.316
.358
.407
.059
.647
.079
.465
~lovers, 1968-69.
WOmen
Stayers
MOvers
.107
.318
.,.297
.099
.276
.091
.240
.078
3t ayers
.172
.149
.129
.097
60
.209
.180
.156
.067
.05i
.049
at the ratios of women’s
.361
.318
.246
.187
.143
.107
h= 12 years OF
For example,
11 percent
incre=e
chance of separation
chance in 1980 -8.l—a tw~fold
to 8 percent
it is apparent
during
that
The only exception
of cument tenure during the period
a rapid sectiar
men faced a 17 percent
and a 32 percent
from
probablliti=,
l~s Ukely to leave their jobs in the next six months.
dso shows that men have undergone
dropped
‘
Movers
.407
.072
.054
.040
to men’s job separation
workers with more than three ye=s
in 1968-69
Men
ratio.
w,omen are generally
=ong
1980-81
WOmen/Mm
Men
.341
.361
srker is whtte, marri
“A modd
schooh”g, and hv- in an SMSA
Glancing
Periods.
r, For Stayers anc
36
48
the femde-mde
\VOrkers”.
1976-77
W’omen/Men
Men
.661
.706
.808
.927
.172
.149
the Next Six Months.
fOrkfOdal
in the probability
1968-69.
is
Table 10
of job separation.
in the first six mmrths of a new job
incr-e.
For women,
the same period .14 In 1968-69,
I{The ~c~ar d=re== im the prObabifitiS fo, women k sbghtly undemtatd
the probability
women
because the probabfliti=
faced
we~
a
CV~-
nton the unemployment rate
uated at the average unemployment rate for the entire period. The estimated coefiti.
and the latter w=
is -.o21 for women, implying that the hmard d=rem
u the unemployment rate incre==,
about four percentage points below (above) average in 196&69 (1980-81). Thh k not a problem for men since their
coeffi.ie”t OD the unemployment rate is very CIO% to zero (0.002).
27
probability
of job separation
in the first six months
m that of men. This retie is 34 percent
TO compxe
sepmation
the behavior
than female movers.
for men than far women,
effect
to t~e
trmsitory
1968.69.
hold.
thm
Although
it is unhkely
For this reason,
to unobserved
qutities,
so it is not ~t”dly
to be vindicated
be discour~ing
have succesdve
labor force participation
TO demonstrate
rat-,
sumive
long enough
movem
are somewhat
of mde
within eafi
gender.
re~on,
conttins
probabilities
standpoint.
employees;
they a
cohorts
with
for this
more
thm their male counterparts,
Of course, the stayer d~ignation
If employer:
refers
simply
then they me
w we lewned
from Table
9, 54 perwnt
of the women.
Although
this news ,may
take mmfort
of women incre~ed
in the knowledge
their education
but they have altered their turnover
of the h~ard
behavior
that times =e
atttinrnent
~d
M weU.
birth
Our early cohort consists of women who were born in the period
1944-
consists of women born in the period
Thre-yeu
observations
model for m “early”
1952-54.
sufficieritly large saples;
to be zero at the first ob=rntion.
on these workers while the r-pondents
sizes for the men are 853 ad
are betwen
cohort and 1,019 in the late cohort;
1,438. Summmy
in Tables A2 and A3 in the Appendix.
28
ad
For the men, we use everyone
windows are used to mtinttin
we no longer require prier ~perience
31. There are 788 women in the e~ly
smples
f~ta
a “late”
born in 1942-44 and 1950-52.
md
panel of Table 10.
and Hope that they prove to be stayers,
this, we prwent ~timates
46. The late cohort
smples
that jobs
with only 42 percent
to female non-quitters,
Not ody
for the sme
of job
drops mnsiderably
possible to identify such workers.
by their mde
of the men me stayers compared
cohort
analogous
in the bottom
that movers’ jobs till
ha~re lower separation
workers On the basis of obser}.ables
&mging.
probabilitia
in 1980-81.
the discrete h~ard
we conclude
they may be a better bet from an employer’s
more hkely
m high
those of female movers.
Since female stayers generdy
fire
These me reported
62 percent
reveal that, in the first two years of a new job, mde movers face higher prObabiEties
of job separation
taure
in 1976-77 md 25 percent
of movers to that of stayers, we compute
for movers for the period
The estimate
of a new job that w= ody
The
the ages of 24
the corr-pending
statistics for the four samples are pr+ented
Table 11: Estimates for Gompertz Hazmd Model
Correction for Unobserved Heterogeneity
(SAMPLE
“Early”
ad
‘Late”
Birth COhOrts Of Men and WOmen, from Ages 24 to 31)
Wo
Normal
Normal
20 ef.
.002
Stat.
.02
..049
-.87
-.022
-.192
-.34
-1,23
..077
..191
-1.28
-1.55
Variable
COef.
NONWHITE
MAR~ED
WEDS
DIVORCES
BIRTH
NEWBORN
1944-46
1950-52
1944-42
stat.
COef.
-.111
1952-54
Normal
Stat.
COef.
-1.84
-.023
-.42
2.68
-1.80
-.061
-.020
-1.05
-.14
-1.33
.249
.040
2.72
.19
.217
.316
.480
1.31
4.32
3.30
.184
-.021
-.156
2.65
-.28
-1.71
.028
-.239
.44
-3.77
-4.39
-.159
-.098
1.152
.451
-.432
.278
.549
.515
-1.19
-8.44
-8.76
6.42
-5.64
3.60
5.88
3.10
-1.26
1.95
-2.58
5.57
-1.05
-.54
.119
.86
-.150
—
—
-2.17
—
—
-.002
-.02
scHooL=13-15
SCHOOL= 16+
PRIOREXP
TIMEWORKING
INVOLUNTARY
WAGE
-.061
-.391
-.043
-.401
.006
-.230
-:204
..178
-2.94
-2.29
-4.62
-2.91
-3.55
.11
w3.34
-5.31
-1.40
-5.37
4.18
-2.84
PARTTIME
UNION
CONSTRUCTION
TRANSPORTATION
TRADE
.901
-.635
.486
9.49
-7.05
5.80
-,518
-.018
-.491
.261
-.193
.823
-.492
.546
-.320
‘.106
-2.84
1.60
-.524
.215
-.263
.146
PUBADMIN
SMS.4
SOUTH
UNEMPLOYMENT
Y6667
Y6869
-.464
-.056
-.085
-.120
-.613
-.300
.228
-.424
——
——
——
-3.18
-.84
-1.41
-3.57
-3.62
-2.01
2.02
-3.S8
-.412
-.074
-4.76
3.61
-3.54
-1.33
-2.41
-8.09
—
.852
.762
.699
-.418
—
—
5.01
6.41
5.32
-3.87
—
—
—
—
—
—
—
—
-.018
-9.89
PRESCHOOL
SCHOOL=12
Y7071
Y7273
Y7475
Y7677
Y7879
Y8081
Y8283
Y6485
L(6)
a
O*
Log Mkefihood
——
—
—
-.65
—
—
-.024
-12.58
.1.748
-5.64
-.777
-3.27
1.188
13.73
-6871.740
-.126
-.220
——
—
——
—
——
.535
.053
.407
——
——
8.18
-7.00
7.02
—
7.52
.62
3.54
-.019
-13.00
-.437
-1.31
1.22
.586
-4.54
-.837
-7711.512
29
Normal
Stat.
.190
-.313
-.249
-3.65
-.113
-.274
——
——
EN
-.372
.325
-.067
.242
—
7.21
—
-14.71
-.41
11.42
1.587
-5481.353
4.962
..054
-.371
-.421
-.059
-.28i
.098
-.351
.412
-.087
.218
-4.24
-4.61
-3.24
1.18
-5.74
6.56
-1.17
-.386
.004
.062
.216
—
—
.96
.-.69
3.61
-2.46
.08
1.27
4.92
—
—
—
—
—
—
—
—
—
—
-.127
.242
-.555
-.192
.019
2.707
-.015
3.707
—
—
-5.75
-2.43
.17,
-5.13
-10.85
-9.81
—
—
-5353.675
The cohort
women
~e
compmisons
separated
the early cohort,
significant.
ad
by mdy eight years,
the coefficients
ily obligations
dso
h=ard
l~t
u
educational
goes from positive
behavior
WORKING,
BIRTH
ad
While
observed
Among
whiti
fmily
heterogeneity
the=
increases.
fwtor
The bi~est
l-s
impmt
whatsoever.
are movem
negative
the coefficients
thm
NEWBORN,
they are known at the time of hire.
me, by their ve~
nature, short-~ved
eventi.
tiffwenc=
and TIkfE-
than the early cohort,
dso virtu~y
are movers and their h==d
can be demed
h~md
md in
rate, un-
to the early cohort.
disappem
by a
movers.
predicting
The importmt
but, with the exception
Furthermore,
over time
rate increw-
shmdd have no difficulty
This is in mmked
30
SMSA, SOUTH,
rate of the ,movers is 4.9 times =
heterogeneity
rate are not only observable
over time in
The inter-cohort
wi~ quit when they look at a sample similar to the late cohort of women.
h~ard
on PRI-
for the early cohort,
on the ymmger women’s
by Table 11 is that employers
to s&Oohng
over time.
and the h=ard
of 3.3, while only 17 percent of the late cohort
nmts of these women’s
events of BIRTH
but the late cohort’s
This is in marked contr=t
37 percent
md
that fm-
changes are in schoofing
effects on the late cohort
For
to a mud
With these &mgffi
signs for men and women.
of
si~ificant,
reve~
On only four variables—BIRTH,
have a diminished
the early cohort,
The message d~vemd
Interestingly,
insignific~t).
(ad
high u for the stayers. The effect of unobserved
for the men. Among
dummies,
to negative
both of which become
61 percent
that remtins
differ in thtir response
for the late cohort
h= no impact
women,
dso
positive
but they am cofined
by the s&OOfing
less pronounced.
obligations
aTe ~
~timates
are less negative
have larger (negative)
PRIOREXP,
Th=e
quit rates,
late cohorts
them is quite remarkable.
and PRESCHOOL
effect.
The cohorts
attainment
different
men are @nsiderably
between
md
for six years, but the combined
of women, the coefficients
UNEMPLOYMENT—have
-ong
women’s
the emly
one of these thee
positive
is unflected
and TIMEWORKING
the turnover
lmge,
for only one year.
Although
BIRTH
is the ody
to tiect
cohort
11.
the difference
a pr~chOOIer is mound
of the emly
while UNION
ad
~,
falls shmply
OREXP
a ve~
have not ce=ed
NEWBORN
the h~~d
BIRTH
registers
shorter period—titer
in Table
on NIARRIED,
For the late cohofi,
NEWBORN.
md
me presented
determi-
of BIRTH ad
the two ‘unpredictable”
contr~t
who
to the fmily
factors
chmxter-
istics that rtise the h~md
which tend to endure
of the emly mhort
for years at a time.
(MARRIED,
BIRTH
From an employer’s
md
point
PRESCHOOL),
of \,iew, single women
women who are likely to give birth in the near future appear to be ristier prospects
counterparts.
HO\vever, women
who me Ekely to have completed
be safe bets, e~,en when their children
h
addition
dso conclude
to cltiming
that employers
the non-qtitters
turnover,
some of
and
than their mde
their desired fertifity
app~r
to
are still quite young.
that non-quitters
are identifiable
in the late cohort
of women,
we cm
would be unwise to favor men over women w,hen attempting
from tkis pOOl of workers.
A yourig worker of either gender
is prone
to pick
to future
of course,
but the abifity to predict turnover do= riot incre~e
when we confine our
*
sample to men. In fact. it may actua~y decre~e,
since a smau proportion of the men are movers
for rewons
saple
that cannot be observed.
Obser\,e at the time of hire, i.e., we omitted
PRESCHOOL
nriable
titer contro~lng
than for men.
saple
we ~timated
of men and women (using workers born in 1950-52)
employers
dummy
Furthermore,
and the correction
for unobserved
equal to one for women
for obsemables,
By contr=t,
the hazard model for a combined
and controlled
\VEDS, DIVORCES,
heterogeneity.
is .0.19 with a normal
the hazard rate of job separation
31
BIRTH,
The estimated
statistic
that
NEJVB ORN,
coefficient
equal to .048.
is 17 percent
the gender effect is zero when we peflorm
of men and women born in 1944-46.
only for vmiabl=
On a
That is,
lower for women
the same experiment
on a
Section
k
the previous
being undertaken.
by mother
Estimates
5:
section,
we exafined
Jobs cotid
heterogeneity
in individrrds’
about the comparative
job separatimrs
end tither
job, unemployment,
for Selected
voluntarily
non-employment,
commitments
job shopping
model for a vmiety of sub-smples.
trmrsition
undergone
Entry
ad
employed
they cotid
be fo~owed
spe~. In addition,
we i~ored
In order to augment our conclusions
of men and women, we now re-estimate
First, we 100k at five separate l~or
or involuntary.
by continuously
regardless
Finally,
our h~md
market entry cohofis.
We
of the type of spe~ that is entered
by another job, regardless
we restrict oursdves
of whether the
to the job-twjob
trmsitions
workers,
Cohorts
We focus
on workers
questions.
wend
or involuntarily,
a sample of jobs that are fo~owed
is voluntary
regard to the type of trmsition
or an unidentified
then focus on a sample Of jobs that end voluntarily,
into. We dso ex~ine
without
to the labor force.
behavior
Sub-Samples
careers began in a specified
The first is whether
is whether
estimated
sdected
whose
the behavior
our h~ard
workers whose
model
of men ad
entry cohorts
begmr in 1966-67,
wer
time.
to adtiess
behavior,
for each gender.
i4-75;
two
the
this, we
For the men, we
for the women,
Workers in each cohort
workers whose careers begin
md
TO accomphsh
68-69, 70-71, 72-73 md
70-71, 72-73, 74-75 and 75-76.1s
for the first six years of their careers:
year in order
differ in their t~rnover
women convwges
for five successive
meers
the entry years me 1968.69,
succe~i ve entry cohofis
calendar
are fo~owed
in 1966-67 ~e”followed
rrntfl
1971, etc.
Estimates
dramatic
shink
ad
for the five entry cohorts are presented in Tables 12 ad
numbers in these tables are in the bottom
considerably
with each sumessive cohort.
63 workers (2,373 and 833 observations),
15T~=
fir.t
but who sti
entry
wre
cohort
for
both
~ender$
d=
two rows, whi~
indicate
The third female cohort—which
workiri whose care=
in their fist jobs.
32
startd
the most
that the smple
The fourth and fifth mde cohorts conttin
r~pectively.
incl.d~
13. Unfortunately,
sizes
only 170
entered
prior ~ the first”interview
.—
in 1972-73,
the same years as the fourth male cohort—h-
while the next two cohorts
conttin
161 and 109 w,orkers, respectively.
us to discount
the later cohorts—and,
stops
with the fourth
abruptly
Nevertheless,
cohort.
in fact,
cohorts.
ad
on MARRIED,
WEDS
The coefficients
TIMEWORKING
For the women,
dl become
pronounced
in the c~e
with ewh
of WAGE.
one of the few with a non-zero
our attention
fitited
Utitation
mounts
relatively
other in two important
Voluntary
Table
negative
for
pattern:
the effects of
f=bion.
on PRIOREXP,
TIhfE\VORKING,
and }i’AGE
although. the pattern
is the most
variable
that shows a pattern
(and
the presence of w infant raises
Because we must restrict
across genders.
Unfortunately,
is that the we
Since W the members
their car~rs
of mde =d
ad
fem~e
schoofing
is too
turnover behavior.
distributions
shift to the
of the full sample were 14”to 24 y-s
later in the survey were relativdy
k other w,ords, the five entry cohorts
demographic
this information
old
younger
differ tigtificantly
from
&mensiOns.
~ansitions
14 presents
thaae smple
wtimates
are reported
NLS respondents
r~pons~
However,
mde
for each gender. there are only two ( 1968-69 and 1970-71)
of these estimates
better educated.
incre~ingly
with each successive entry cohort.
comparison
cohort.
become
is h’EWk ORh’. Apparently,
in the first year, the workers who begs
md/Or
detect
for each sucussive
to -1.33 for the 1970-71 cohort.
The only non-job-related
effect)
a v~d
right writh each succ=sive
that we ca
display an even more pronounced
to allow us to draw inferences about the convergence
Another
=ch
and DIVORCE
of the first three cohorts,
to the first three cohorts
with which we can m&e
pattern
more important
evolve in a non-monotonic
the hazard rate by incre=ing
These sma~ samples compel
every secular
on PRIOREXP
we find that the coefficients
more negative
(7,433 obser~,ations),
cohort,
they go from -0.023 for the 1966-67 cohort
WAGE
almost
we do see that several variables become
The coefficients
the first thr-
only 419 workers
to their
b~ed
in Tables
were repeatedly
l~t
job,
on samples
=ked
thereby
A4 ad
of voluntary
transitions.
A5 in the AppendIx.
why their lmt job ended.
Iearllillg why a subset of jobs
33
Summ=y
As discussed
statistics
for
in Section
3,
We are often able to fin~ thaae
will end.
When
the repofied
Table 12: Estimates
for Gompertz
H~=d
Model
Correction for Unobserved Heterogeneity
(SAMPLES:
F,ve Entry Cohorts of Men)
—.
1966-67
P
NOnwKlte
Mamied
Weds
Divorces
1968-69
~
1970-71
B&
bg
.065
-.188
1.16
-3.59
.031
-.200
.45
-2.71
-4.11
.1.98
-.381
-.700
-3.36
1.85
Birth
-.395
.4&7
-.423
-.563
-5.37
School
PriOrmp
-.033
-.023
-5.39
-3.71
-1,07
.858
-.454
-.014
-.482
-,94
-11.99
-.069
-1.330
TlmewrHng
Involunt=y
-.815
.331
-9.59
4.27
8.37
:.88
-.038
.428
-.734
.285
-.167
.014
-.351
-.152
-.003
-.057
-1.282
-.79
6.71
-8.16
4.28
-1.64
.24
-2.58
.723
.351
-.244
.241
Wage
Pmttime
Union
COnstmctiOn
=asportation
made
Pubtitin
SMSA
South
Unemployment
Yea dummy I
Ye=
dummy
2
7
10g(6)
Q
Oz
Log Ekelihood
Ye= dummy 1
Yew dummy 2
# of Workers
# of Ohs.
1972-73
B“
.094
-.509
-.787
1974-75
Bg
.90
-5.34
-4.48
2.25
-.069
-.386
-.289
-.271
-.26
%
-2.05
-.88
-.50
-2.93
-:370
-3.23
-17.78
.027
-.124
-2.034
2.623
-1.17
.51
-.86
-.426
.374
.429
1.463
-2.342
-.83
.93
.61
1.52
-2.65
-.172
-.526
-2.291
.056
-.279
4.218
-1.003
-1.377
-3.417
-2.078
-.607
-1.56
-1.95
-268
.06
-.81
6.32
-1.82
-1.75
-3.48
-3.45
-.93
-.359
-.481
-.81
-1.18
-.046
-2.053
-.55
-3.42
-.576
-6.32
.255
.3.18
-.172
-131
.238
3.75
-.241
-1.51
-.227
-3.21
.014
.22
.027
1.52
-.029
-.29
-1.61
-.160
-9.64
-.038
-1.655
-6.31
-.888.
-6.22
1.270
.17.67
-5814.019
-1.392
-11.79
.301
2.28
-.192
-.2.45
.665
6.30
.1.083
-7.83
.485
3.40
-.413
-2.52
.033
.36
-.519
-2.09
.043
.41
-.367
-4.43
-.172.
-7.47
-.059
-.48
2.142
9.67
-.083
-14.23
3.510
8.87
.1.110
-7.39
.2.157
-13.45
-2281.705
-.842
-3.20
-.026
-3.24
.11.839
.09
.320
2.39
13.391
.11
-706.247
.568
1.1s
-.016
-1.08
4.271
1.73
.026
.13
-3.866
-6.57
-243.233
Y6869
Y6869
Y7273
Y7273
Y7475
Y7273
Y7475
Y7475
Y7879
1183
18,175
741
10,507
395
5,759
170
2,372
63
.833
:2.90
-.07
-2.94
-10.19
-.507
-5.20
-.041
-18.23
-.969
-3.86
-.154
-.70
1.019
8.84
-8904.015
Y6667
-3.92
-3.44
34
-3.83
2.32
-.913
-.069
-.400
.208
.334
.018
-.502
-.161
-3.56
-.26
-.93
.71
.76
-.303
-.095
-2.918
-1.68
-2.59
-6.08
.08
-1.19
-.88
retion
is health,
f-ily,
pregnancy,
we view the job w ending
job dissatisfaction,
in a \,oluntary transition.
found a better job,
There
school,
are 3,403 such jobs
or other quit,
for the men and
3,418 for the women.
Since Table 14 is hued
should be compared
ad
WAGE
trmsitions;
me roughly
the sme
regardless
the primary
difference
is that PRIOREXP
transitions
vmiables,
we see that
all-transitions
~ it is in deterring
smple.
MARRIED
the hazard.
PWOREXP
TIM Etl’ORKIN
we look
negative,
but h=
(and significant)
and WAGE
G, however,
transition,
at all traditions
determinant
or volunt~
a factor
LooKlng
a smaller
it
TIkfE\VORKING,
is not ,X important
is not m important
had a negative
the women,
of whether
on PRIOREXP,
other types of transitions.
remtins
AIso, SCHOOL
it previously
Among
who report a voluntary
to Table 9. For the men, the coefficients
voluntary
although
on dl workers in the “full sample”
in deterring
at the demographic
magnitude
than in the
of voluntary
transitions,
coefficient.
continue
to have the same negative
is no longer a significant
factor.
Apparently,
effect
mr
a woman’s
labor market history influences
Surprisingly,
Women
who have pre-school
Although
increwe
her chance of being laid Of or fired, but. not her propensity to quit.
.
we see that the coefficients On WEDS and NEM’B ORN me no longer signifimnt.
this latter
the h==d
aged children,
finding
prompts
on the other hand, are more fikely to quit their jobs.
is unsu~rising,
us to =k
whether
tbe fact that WEDS
our memure
ad
of voluntary
NEWBORN
transitions
Not mdy might workers provide insincere responses to questions about the re~ons
sepmations,
but the difference
between a quit and a layoff is conceptu~ly
because so few jobs can be *sociated
tith
a nonrepresentative
dtia
qutity
Job-to-Job
We cm
smple.
with a re=on
Although
behind their job
indistinct.
dissolution,
is refiable.
Furthermore,
we may be left
we do not entirely distrust the results in Table 14, the
is not high enough to wamant fmther
analysis.
~ansitions
view our sample of job-to-job
transitions
with considerably
Section 3, this sample consists of dl job separations
by a new job.
for their eventual
do not
That is, we elitinate
virtually
d
As noted in
that are followed less than three months later
job-t~unemployment
35
more confidence.
and job-tenon-employment
Table 13
Estimates
Correction
(SAMPLES:
1968-69”
Nonwhite
Mmied
Weds
Divorces
Birth
Newborn
Preschool
School
PtiOrexp
TlmeX\,Orking
Involuntuy
M7age
Parttime
Union
Construction
Transportation
Trade
Pubtitin
SMSA
South
Unemployment
Ye= dummy I
Ydummy 2
7
10g(6)
o
82
LOS Hk&hood
Ye~
Ye=
# of
# of
dummy 1
dummy 2
WOrkers
Ohs.
for Gompertz
for Unobserved
Rve Ent~
1970-il
+
P&
.077
.280
1.16
4.50
.086
.014
.003
-.556
.263
.03
-2.98
3.43
3.91
-.79
-1.18
-.399
-.391
-.385
.680.
-.285
.000
-3.79
-1.64
-4.01
-8.96
-5.32
7.96
-10.48
-.439
-.674
-9.61
-5.71
.643
-:780
6.19
-10.80
5.86
12.85
2.48
-3.42
4.44
-2.44
2.25
1.69
-2.22
-4.60
1.22
.365
.823
.281
-.922
.297
5.01
12.66
1.29
-5.60
4.27
-2.76
.25
.72
2.82
1.71
-.26
-.343
.142
.102
-.03i
-.431
.094
1.16
.20
5.11
-3.14
.03
-.515
.019
.050
.038
.142
-.044
Cohorts
$
.162
.164
-.176
1.47
1.73
-1.14
.436
-.203
.902
1.57
-1.71
4.79
-.364
-2.75
1.02
.027
-.359 .-6.04
-.622
-3.65
1.86
.304
-6.89
-.805
.482
4.67
.786
7.11
2.08
.597
1.150
.539
.080
.054
.017
.076
-.389
1.09i
-3.35
4.89
3.44
.55
.18
4.09
-1.76
-7.78
-.030
-7.20
.2.693
-10.52
.103
1.35
2.107
17.42
-4795.413
Y6667
Y6869
Y6869
Yi2i3
Y7273
Y7475
964
13,869
725
419
7,433
36
1976-77
1974-75
b
-.030
-14.55
.1.565
-5.75
-.137
-1.21
1.470
16.42
-6613.047
10,811
Model
of Women)
1972-73
b
.433
-.055
-.017
-.259
-.560
.581
-.651
.382
.835
.449
-.468
.280
H==d
Heterogeneity
-.016
-2.97
3.789
-6.90
.24
.034
2.058
9.64
-2640.479
B&
-.li4
-.228
.325
.419
.071
.688
.135
-.042
-.275
-.335
.753
.248
.776
-.142
.393
.185
.190
-.114
.192
.205
-.150
.221
-.483
-.79
-1.11
1.09
.78
.28
1.72
.42
-.72
-1.81
-.66
1.75
1.06
3.72
-.44
.39
.36
.71
-.30
.9i
1.06
-2.49
.80
-1.53
-.030
-3.10
.1.374
-1.09
.13
1.782
.200
.50
-70i.034
Yi2i3
Yi4i5
161
2,647
B
&
.184
.024
.59
.07
2.97
1.630
19.419
.239
-1.139
.278
.025
-7.75
.73
-.545
-2.159
-1.4s
.53
.27
-2.85
-2.i5
-.902
-.33i
1.427
.093
1389
-1.31
-1.20
4.65
.25
1.42
19.251
1.107
-.232
-.543
-.405
-.282
-.295
2.271
-8.li
1.77
-.39
.l.~~
-.001
1.631
-.15
1.16
-1.06
-4.09
-.81
4.39
-.55
-.086
-8.46
-3.845
409.085
Y7475
Y7879
109
1,693
Table 14: Estimates
Correction
(SAMPLE
for GOmpertz
for Unobserved
Haz~d
Jobs Ending in a Voluntary
hI EN
Vmiable
NONWHITE
MARRIED
WEDS
DIVORCES
BIRTH
KE\YBORY
PRESCHOOL
Transition)
WIOMEN
Normal
Statistic
Inefficient
Llodel
Heterogeneity
~oefficient
.163
-.125
1.99
-1.90
-.009
-.263
.212
-2.22
-.131
-.101
-.336
—
.99
-3.99
—
—
,017
Normal
Statistic
-.14
.29
-1.33
-.75
-.58
SCHOOL
PRIOREXP
TIMEWORKING
-.006
-.48
-.047
.161
.150
-.022
-.059
.444
-4.27
4.53
-.0i8
.072
-5.68
.77
INVOLUNTARY
WAGE
P.4RTTIhlE
.310
-.182
.584
-.135
.331
-.165
.243
.435
-.495
.296
.721
.472
-.566
.270
-.216
.088
5.12
-7.34
4.74
.016
.053
3.65
-2.85
6.89
-1.42
3.77
-1.14
3.71
-2.16
2.65
.25
3.48
Y7475
Y7677
Y7879
Y8081
Y8283
.617
.853
-.083
-1.438
.046
.862
1.272
—
3.83
4.75
-.42
-6.17
.21
4.02
4.82
—
Lg(t)
a
-.009
-1.390
.263
-5.71
-5.82
.054
.053
-.547
-.639
,-.967
-.864
.153
.634
.002
-.705
-1.208
UNION
CONSTRUCTION
TRANSPORTATION
TRADE
PUBADMIN
Sh$SA
souT1i
UNEMPLOYMENT
Y6869
Y7071
Y7273
62
Log Ekefihood
—
-.362
.173
3.20
-18.24
-1.930
-6212.378
37
-.004
-.023
1.28
2.18
-1.52
10.43
2.57
-3.48
4.37
-1.37
1.36
-.08
-1,73
.17
,16
-1.68
-1.86
-2.75
-2.48
.41
1.51
1.12
-1.91
-7.89
-12.91
-1.533
-6323.322
trmsitions.
The estimates
coefficients
for these. s-Pi=,
On a number
which appear ii Table 15, reveal several contr~ts.
of key t.ariables
go to zero.
-0.205 for the men in the full (dl
transitions)
we saw in Table 9, the coefficient
on WEDS
neithw
coefficient
for both
genders,
PRIOREXP
women.
rem~ns
(Table
is now smaller
effect
unemployed,
factors
in explaining
for women.
d
are far less Ukdy to chmge
a negative
drmatic~y
md
relatively
mpect
recently
tildren
tight
that job-to-job
in explaining
movements
the hward
for
job-to-job
to accept
transitions
lowers the h=ud
a new job.
than they were
for men,
but not
of roughly -0.i for both genders; that is, both men ~d
women
coefficient
for women,
We d~o see that DI\rORCE
whi]e the effect of pRESCHOOL
women to be more Ekely to accept
be constrained
goes to zero
rates of the full s~pls.
to what we saw in Table 9. Both results ~e somewhat
divorced
but
Hkely to leave the labor force
propensity
jobs around the time of a new birth.
significant
for women,
no longer rtisw
make workers more or l~s
In Table, 9, we sa~v that BIRTH
h’ow w,e see a coefficient
Relati>,e to v,hat
and larger
on the huard
but they haie no effect on a worker’s
tmsitions.
which W*
md TIMEWORKING
no longer lowers the hazard for men, while UNION
W of th~e
on MA RMED,
9), is now zero.
for men
effect of SCHOOL
The
Other variables prove to be more important
h=
sample
although. each had a negative
Apparently,
or become
si~ificmt.
The coefficient
First, the
in their ablfity to undertake
are generated
by a smder
a new job,
a joh hunt.
set of covariates
now.
incre=e~
surprising,
since we
tihile women with young
Overall,
thm
however,
ae
other
we see
types” of
traditions.
Continuously
In focusing
Employed,
on job-tejob
their overall commitment
Workers
trmsitions,
we effectively
to the labor force”; in some sense, we were looking
job shopping
‘at the moment.,,
job shoppers
by virtue of their demonstrated
workers with a demonstrated
a new job.
TO identify
We now 100k at a smple
commitment
these workers,
commitment
we me~ured
regardless
of
at workers who were
Of w~rkeri””who we fikely to be ‘exnestto the labor force.
Though,
to the labor force need not be constatly
38
●
I.ooked at dl types of workers,
the percent
of their observed
of course,
shopping
for
time that is
T~ble 15: Estimates for GOmpertz Hazmd Model
Correction for UnObser\.ed Heterogeneity
(SAMPLE
Job-t&Job
~msitions)
MEN
Normal
Vmiable
NOXWHITE
MARRIED
coefficient
Statistic
-.101
.091
-1.24
1.27
WEDS
DIVORCES
-.271
-.121
-1.61
BIRTII
-.733
—
-.38
—
-6.13
—
—
SCHOOL
PRIOREXP
TIhfE\!’ORKING
-.002
-.13
-.003
.022
INVOLUNTARY
WAGE
PA RTTIME
.160
-.244
.224
UNIOY
CONSTRUCTION
-.571
.214
-.18
.23
1.82
-3.41
2.31
-3.96
2.43
TRANSPORTATION
TRADE
PUBADMIN
SMSA
-.018
.072
-.654
.013
-.Oli
NE\VBORN
PRESCHOOL
SOUTII
UNEMPLOYMENT
Y6869
Y7071
Y7273
Y7475
Y7677
Y7879
Y8081
Y8283
70g(J)
a
e=
Log hkehhood
.028
-.326
-1.662
-.710
1.03
.26
-1.59
-3.85
-3.60
-.64
3.24
1.29
-4.12
.15
9.26
-6.34
3.06
11.55
2.48
-.10
.92
-.337
.li4’
-1.09
1.55
-3.52
.17
-.24
-1.85
-.368
-.188
.183
.251
—
-1.59
-1.81
1.58
10.30
—
-.317
-1.61
.01
-7.25
-4.90
-5.06
-5.68
.49
-2.34
-10.59
-1.34
18.41
7,62
11.51
2.12
1.71
3.91
4.96
7.42
—
-.024
-2.378
-1.120
2.317
-9.66
-9.49
-9.67
13.26
39
Statistic
.125
-.144
.441.
.032
-.125
.027
1.225
-.539
.365
1.213
.708
-.035
1.253
2.193
.495
.437
1.041
1.366
2.353
—
-3709.703
~efficient
.003
-2.410
-1.489
-2.014
-2.512
.225
-.007
-4.213
-.167
1.988
-1848.164
known to be spent w,orkin~ anyone with a number
percent)
is considered
comtitted
to the work force thm
Looking
at Table
frdl s-pie.
&ect
on continuously
most
percent
of the ffl
tiong
it h=
9, but
employed
of th~e
we see that it h=
becoting
~MEWORKING
for selecting
we expect
higher numbered
jobs.
in TIM EWORKING
arises horn the difference
of turnover.
among continuously
employed
Presumably,
employed,
employed
jobs
workers,
This is m inter~ting
women.
employed
workers,
second
~d
the vmiable
subsequent
across second
and
this effect wodd
disappear
‘career
TIMEWORKING
the rest.
We
is no longer
finding since it impli~
a
that first jobs
effwt
is BIRTH.
The coefficient
for women incre=es horn zero
.,
if we cmdd 100k at workers who w~e trtiy
schedule,
Or even part-time
dso inme~es
women”
up the jOb-tw OLF movements
An alternative
upon giving birth to a child.
on SCHOOL
ad
workers.
since it is fikely to be piting
hours, or jobs on a tied
first jobs
to Table 9, but the coefficient
rather than job changes.
that we~ educated
in
employed
workers,
in TIMEWORKJNG
that now show a more pronounced
for men does not change relative
The coefficient
men and 59
on PRESCHOOL
of continuously
in its value betwen
higher numbered jobs for the=
the vuiables
&OOse to chmge
no effect
Since the vmiable is set equal to zero for the first job, most of the variance
determinant
Chief mong
in the
an insignificant
employed
coefficient
of continuously
that there is fittle mriace
significant
in our smple
on the turnov~
the s~pl~
for the samples, of continuously
ue no shorter thm
evident
effect on men.
find that,
mntinuonsly
We saw a positive
is Ukely to have a value clo~e to one for most workem’
Therefore,
to 0.199.
but it h~
of the continuously
married deters turnoveq
nO effect On w~men and a smder
the criterion
of several effects that ae
it is likely that marital status b=
69 percent
no effect
(88
man in our smnple..
men. TO some de~ee
men are married:
distribution
these are workers who ~e more
for men in the full s~ple,
sample of men are married.
now
More accurately,
is the medim
lowers the hazard
the full smple,
~lven
jobs.
mnpioyed.
16, we agtin see the disappearance
MARRIED
because
Table
“c6ntinu0usly”
abo~,e the median of the mde
explanation
Women
for this effect is that women
may favor jobs
that d-red
fewer
jobs.
for women,
from -0.026 to 0.039.
~e succeaaful job shoppe~.
40
that remtin
This may reflect
We dso find that the coefficients
Table 16: Estimates
for GOmpertz
H==d
Model
Correction for Unobserved Heterogeneity
(SAMPLE
Continuously
Employed Workers)
WOMEN
MEN
Nomd
Normal
:oeficient
-.083
Statistic
-1.57
~oefficient
-.142
-1.33
-.088
-.109
DIVORCES
-.061
-.292
.211
BIRTH
NEtVBORX
-.376
—
-6.66
—
—
—
-.025
-.102
.016
.420
-.207
.680
-.408
-3.15
-10.38
.27
6.82
-4.60
11.41
Vmiable
NONWHITE
MARRIED
WEDS
PRESCHOOL
SCHOOL
PRIOREXP
TIM E\VORKING
INVOLUNTARY
WAGE
PARTTIME
UNION
CONSTRUCTION
TRANSPORTATION
TRADE
PUBADMIN
SMSA
SOUTH
UNEMPLOYMENT
Y6869
Y7071
Y?273
Y7475
Y?677
Y7879
Y8081
Y8283
7
log(d)
Q
82
Log Ukehhood
-3.67
1.49
.326
-.436
.173
-6.40
5.71
-4.70
3.81
-.577
-.100
-5.29
-2.45
-.002
-.028
.507
.877
.s75
-.04
-2.58
6.23
8.99
1.70
2.57
4.58
6.26
4.17
—
.315
:602
.802
.694
—
-.022
-2.335
-1.718
.942
-13647.154
41
-21.41
-16.54
-5.65
13.57
Statistic
-1.66
-1.20
-.099
.199
.286
-.93
-.46
1.77
1.46
-.149
.039
1.50
1.88
-.060
-.245
.334
-.649
.580
.263
.407
-3.44
-1.88
3.05
-6.95
6.36
3.17
1.75
-3.77
-.581
.296
-.410
.181
3.36
-2.82
2.19
.015
.055
-1:178
-1.222
-1.116
.21
3.53
-3.06
-3.21
-2.93
-1.849
-1.805
-1.855
-1.891
-2.178
-.015
-2.440
-1.115
-4.64
-4.50
-4.60
1.237
-5585.833
-4.45
-4.77
-9.20
-5.54
-3.93
12.75
on PRIOREXP
employed
are 1=s negative
workers mtinttin
for both genders.
a relatively
This may point
high (and freneficid)
to the fact that continummly
rate of turno~,er for a longer period
of time than does the fuU smple.
The fial
=amined
r~tit
that emerges
the fall. sample,
of 2.8 for the men
on continuously
conclntions
=d
employed
from
Table
we learned that unobserved
3.4 for the women.
workers:
the corresponding
is virtu~y
sttiistics
about movers and stayers are now significantly
continuously
employed
is still more fikely to “draw”
men, -ch
pOOl conttins
to predict
which continuously
workers,
a mover from among
employed
,.
42
unchmged
Among
d
we
by a factor
when
we focus
However,
our
workers, we found
for unobserved
are 16 and 28. Although
In other words, mployers
When
rtises the h~md
are movers
a pool of women
workers will quit.
heterogeneity.
me 2.6 and 3.5.
different.
of the women
these percent~es
very few movers.
unobserved
heterogeneity
TMs r~ult
that 46 percent “of the men arid 58 percent’
tiong
16 concerns
~
than from mong
cm readily
re=ons.
emPlOYer
a pool
use observable
of
Section
In this study,
we estimated
hazard
6:
models
order to learn how they differ in their turnover
questions
posed in the introductory
DO young
model
ad
can obseme
negative
roughly
md
sample
of men and women,
overd
hkely thm
TIUS indicates
counterpxts.
found
only for characteristics
on FE~fALE
men to be mo~,ers (for unobserf,ed
of combined
reasons).
man to be a quitter.
For the late cohort,
women have lower quit rates thm
DO young
Mthough
ad
women
(potential)
our hazard models.
men and women.in
a more pronounced
hypothesis
quit primarily
job matching
to have b~ies,
our fu~ samples.
hypothesis
by including
We find that increties
degrw
however,
such me=ures
When
at successi~,e labor m=ket
we focused
are more
that women
unequivocally
dependence,
entry cohorts,
mrer time; ~ong
on ~roluntary and job-twjob
43
engage
in job
w prior experience
and wages lower the h=md
E\,en though the estimates
of negative duration
even more important
women
s
that
shopping?
to a rigorous test, we control for match qutity
in experience
that the labor force consists of significmt
have become
of both movers
is mor,e likely than an
we conclude
or do they dso
reveal that the h~ard
our estimates
and wages in
rates of both
for men h=
are consistent
numbers of ‘carwr-oriented”
the early part of their careers to shop for a durable employment
we looked
have lower quit
therefore,
woman
have
men.
we do not put the job shopping
previous
actua~y
birth cohorts,
We condudp,
that the
in the early cohort
These overall quit rates reflect the activities
that, in a saple
a haz=d
is zero for the early cohort
that men and women
a grmrp have a lower quit rate than men, but that an indi~,idu~
individud
to the three
When we estimate
quit rate, but that w,omen in the later cohort
M’e ha~re dso
men and women in
We cm now pro}fide aswers
controlling
at the time of hire, the coefficient
thtir mde
stayers.
behavior.
of young
section.
for the late cohort.
the sae
rates tha
for various samples
women separate from their jobs more often than men?
for a combined
employer
Conclusions
relationship.
with the
women who use
Furthermore,
when
we found that the effects of prior experience
women, the same is true for the current wage.
traditions,
,we found
that the presence
of a
newborn
dther
has no effect
type
~plOyed
of transition,
hov,ever.
=e mOre ~kely tO l~ve
is not influenced
women
on the h=ard
md
h
Women
addition,
with a pres&OOler
we learned
are more Hkely to mde
that women
who
are continuously
their jOb ~~rhenthey give birth, but the turnover
by either ne~.born
the late cohort
rate.
Or preschool-age
of women
children.
of these women
Only among continuously
is there the su~estion
.employed
that the birth of a baby induces
tmnover.
Given
that employers
can only observe
more difficult for them to identify
learned (U did He&mm
our full smple
non-qtitters
reveals considerable
unobserved
quit for re~ons
it is harder to identify
among
were to hire a man and a womm
that cannot
be observed
becomes
detetinants
of women’s
turnover
a
insignificant
child and the act of becoming
but neitha
represents
a long-term
can be identified
We re~h
fact, much
with sitilar
This fi.ndiig,
obsemable
coupled
with
ez ante, leads us to conclude
that
a sitilar
married.
deterrent
equfly
conclusion
factor ~ong
that may not be hewn
a newborn
non-qtitters
h p~ticul~,
the women—in
we focus on the late cohort of wo~kers, ho,<.ever, our conclusion
hetemgentity
women, ad
Atered.
the only impoti~t
at the time df hire ~e the presence
Both factors
tO job st&llity.
weH song
is dramatically
of
mise the hazard rate of women,
Among
this cohort,
we conclude
that
the men and the “women.
when we focus
on continuously
employed
workes.
k
these
-
16 percent of the men and 28 percent of the women cm be termed movers for unobservable
or unmeuured
re~ons.
wmdd be unhkely
women
heterogeneity
women me Ake.
We have
female non. quitt ers.
Unobserved
smples,
at the time of hire, is it
among the women than among the men?
the woman would be more ~kely to emerge x a quitter.
the fact that women
When
of chmacteristics
and wIIfis (1977) and others) that not d
more than among the men. If an employer
&aracteristics,’
a handful
to discover
who become
birth Me.
identified
married
We conclude
reg~dless
Clearly,
an employer
who mndomly
that the worker is a mover.
hires a
Furthermore,
are not more hkely to leave their jobs,
that, among continuously
of gender.
44
employed
employee
of either gender
continuously
&though
women
workers, non-quitters
-
employed
who give
be readily
.
References
Barnes,
William
F. and Ethel
Human Resoum69(FdI,1974):
Batiel,
nMdeandF
emdeQuitting~
Joumalof
Ann P. and GeOrge J. Borjas,
‘Wage GrOwthand JOb~rnOver:
An Empiric~AndUniversity Of ChimgO
ysis,” in Sherwin Rosen, editor, Studies in La&r Markets. Chicago
Press (1981): 65-90
Becker,
Blau,
Jones,
“Dlfferencwi
439-451.
Gary
S.,
Wancine
Workers:
Human
Capital.
D. and Lawrence
Industrial and Lakr
New York:
Columbia
University
Press, 1975.
M. Kahn,
“Rxe
and Sex Dlfferenc~
in Quits by Young
Relations Ret,ieu, 34 (July, i981 ): 563-577,
COX, Donld
, “Panel Estimates of the Effects of Career Interruptions
Economic lnquir~ 22 (July, 1984): 386-403.
on.tbe Earnings of Women?
Donohue,
John J., “A Comparison of Male-Female Hazard Rates of Young Workers:
School, Program in Civil Llabitity, Working Paper No. 48, August, 1986a.
“Hazards @tes of Young Mde and-Femde Workers-Recent
Developments;
Program in C,vil LiabiHty, Working Paper No. 51, October, 1986b.
~001,
Flinn,
Yde Law
Yde’”Law
Christopher
and James J. Heckman,
“Models for the Analysis of Labor Force Dynamics.” in R. B=mann
and G. Rhodes, eds., Adoances in Emnometrics,
VOI I. Green\vich:
JAI Press (1982): 35-95.
HasKlmOtO,
Masanori,
“Firm-Specific
Human Capital u a Shined ~vestment,”
nomic Reuieu 71 (June, 1981): 475.482.
Heckman,
James J. and Burton
metn’cs 24 (January/February,
‘A
~dels
Singer,
1984a):
‘Econometric
63-132.
Duration An&ysis,n
Method for Minimizing the Imp-t
of Distributional
Assumptions
for Duration Data,” Economettica
52 (March, 1984b): 2il-320.
Amen”can
Jouma[
Eco-
of EmnO-
in Econometric
Heckman,
James J. and Robert
J. Willis,
“A Beta-1ogistic Model for the Analysis of Sequential Labor Force Participation
by Married Women ,“ Joumai of Political Emnomy 85
(February, 19i7):27-58.
Kalbfleisch,
John D. and &ss
L. Prentice,
New York: John Wiley and Sons, 1980.
Light,
~er
The Stattitiml
Analysis
of Failure
Men;
unpublished
Audrey,
“Job Shoppin
and the Wage Growth of Young
sertation, University of C 3 Ifornla
~ at LOS Angeles, 1987.
‘Job Shopping md the Effect of WWe
No. 306, May, 1988.
Li11ar~~8~=
A.,
and
Linda
J. Waite,
in Mde
Meitzen,
Mark E., ‘Differences
Economics 4 (April, 1986): lS1-16i.
Cuts On”Wage Grrrwth,” Stony Brook
‘Children
ad
M=itd
Disruption,”
and Female Job-quitting
45
Time Data.
Behavior:
Ph.D.
Working
PAA Meetings,
Joumal
~s-
April,
Of btir
“
Mincer,
Jacob
and
Human Capital?’
Haim
Ofek,
Journal
“Interrupted
W70rk C=eers:
of Human R=oumes
17 (Wint=,
Depreciation
1982):
=d
Restoration
of
1-24.
“Work Expectations,
Human Capital Accumulation,
and
Sandell,
Steven
and David Shapiro,
the Wages of Young Women.m Journal of Human Raourms
15 (Summer, 1980): 536-53.
Smith,
James
P. and Michael
Journa/ of bbr
Economics
~ubman,
Bld
Topel,
P. Ward,
‘Time-Series
Growth
3 (January, 1985): S59-S90.
Paul J., Jere R. Behrman,
and Robin
C. Sickels,
White Differences,” PAA Meetings, April, 1988.
aAge Specific
fiussell,
James and Toni Richards,
Models Usin the Heckmm-Singer
University, 18 83.
“Job hfobifity
“Correcting
Procedure:
and the Carwrs
Force,”
Death
Robert,
‘Job Mobility and Eartings Growth:
A Reinte~retatiOn
of Hmm
Earnings Functionsfl in Ronald G. Ehrenberg, editor, R=eaxh
in La&r Economi~
8, Part A. Greenwich: JAI Press, 1985.
Topel, Robert and Michael P. Ward,
Eshed paper, Febmuy,
1985.
fima,
in the Female Labor
Capital
Volume
of Young Men:
for Unmemured
Heterogeneity
Office of Population
Researti,
Wta:
unpub-
in H==d
Princetmr
Nancy,
M. Hannan,
imn Journal of Smiolqy
and L. Groeneveld,
aDynmic
Analysis of Event H1stOri~,n Amer84, (1979): 820-854.
.
Reuiew of Econom;cs
and Statist~- 62
Viscusi, W. Kip, “sex Differences in Worker Qtitting,”
(August, 1980): 388-398.
Waite,
Linda
poration,
Ward,
J. and Sue E. Berryman,
Report
Michael
P.,
~;8~d
Forces,”
R-3106-FF,
and
Rad
Santa
‘Women
Monica,
in Nontmditiond
California,
March,
occupations,”
1985.
Hong
W. Tan,
“Tlle Retention
of High- Qufllty
Corporation,
Report R-31 17-MIL, Stita
Monica,
46
Personnel
California,
Wnd
Cor-
in the U.S.
February,
Appendk
,.,
47
Table Al:
(SAMPLE:
Estimates
for Wtibull
H=md
MOdd
Correction for Unobserved Heterogeneity
Workers Whose Careers Have Not Started Or Whose
Jobs Begin or Are in Pr~ress
in tl~e First Ye=
MEN
WOMEN
Normal
Nomd
Vmiable
~oefficient
Statistic
-.000
-.258
-.00
-8.94
-.353
.106
-.285
—
—
-6.75
1.07
-7.19
—
—
-.045
-8.54
-11.83
-12.29
6.57
-10.2i
14.57
-14.21
11.59
NONWHITE
MARRIED
~T~~s
DIVORCES
BIRTH
NEWBORN
PRESCHOOL
SCHOOL
PRIOREXP
TIMEWORKING
Involuntary;
WAGE
PARTTIME
-.072
-.533
.245
-.290
.530
-.5il
.420
-.279
.120
-.348
-.080
-.087
UNION
Constriction
TRANSPORTATION
TRADE
PUBADMIN
SMSA
SOUTH
UN Elf PLO1-lfENT
Y6869
Y7071
Y7273
.005
.974
1.184
.382
Y7475
Y7677
Y7879
Y8081
Y8283
log(k)
k log(p)
.510
.836
.830
.763
—
-.371
-2.023
a
.130
1.233
-30351.003
9.
48
First
of the Sumey)
-4.94
3.92
-5.30
-2.71
-3.15
.65
18.59
18.15
5.61
5.99
9.68
9.96
7.13
—
-28.60
-17.39
1.77
19.68
~oefficiint
-.052
-.007
-.143
-.097
.046
.385
.022
-.017
-.069
-.440
.409
-.579
.398
.444
.359
-.584
.318
-.401
.049
.045
-.015
-.381
-.251
-.573
-1.371
-.817
-1.578
Statistic
-1.57
-.22
-2.71
-1.11
1.19
6.53
.67””””
-2.60
-12.05
-9.25
10.11
-18.62
12.48
13.56
4.12
-7.86
9.90
-6.05
1.59
1.55
-2.46
-3.56
-2.29
-5.20
-11.34
-6.~2-
-12.62
-7.89
-1.051
-5.42
-.781
-15.21
-.200
-10.48
-1.569
2.12
.125
24.62
1.461
-27201.189
Table A2: Sample
(SAMPLE:
“Early”
Means and Standard
and “Late”
Cohorts
Deviations
of 24-31 Year Old Women)
EARLY
Definition
Mean
1 if nonwhite
.298
.457
Variable
NONWHITE
MARRIED
WEDS
DIVORCES
BIRTH
NER7BORK
PRESCHOOL
SCHOOL=
AGE
PRIOREXP
.4i0
.196
.050
.219
1 ifdivorces
during interval
lifchildis
born during interval
.027
.114
.163
.028
.140
.165
.346
1 ifcllild ageone or under
1 if child age six or under
.024
.244
.448
.154
.209
27.794
.031
.244
.434
.181
.219
27.479
4.568
.li4
.429
.496
.385
.414
2.206
Years Of potential
prior experience
btioofactuti
I ifleftl=t
TENURE
WAGE
Years of tenure
Lognf real hourly wage
1 ifwages set by union
1 if \vOrks less than 35 hours per week
1 ifilldustryis
CONSTRUCTION
TRANSPORTATION
TRADE
PUBADMIN
UXEMPLOYMEKT
SOUTH
SMSA
Y7071
Y7273
Yi475
Y7677
Y7879
Y8081
Y8283
Y8485
Number
topotentid
experience
jobin~,nluntarily
.025
3.093
1.743
3.ill
.302
.155
2.824
.49i
6.ii3
.428
.746
.201
.356
.243
1.476
.495
.435
.016
.044
.154
.053
7.703
.39i
.718
.126
.206
.361
.223
1.072
.489
.450
.105
.256
.307
.437
—
—’
.235
.261
.143
—
—
.424
.439
.350
—
—
—
—
—
—
—
—
—
.158
.272
.277
.365
.445
.447
.230
.064
1,019
20.673
.421
.244
.013
.042
.149
.063
.115
Unemployment
rate
1 if fivingin the SOuth
1984-85
—
—
788.
14.865
1.
49
3. Oil
.452
.395
.346
Collslruction,
~riculture,
mining
Trmsp., cml>muniatioll,
utifities
Trtie
Public administration
of individuals
.422
.056
3.764
1.707
3.642
.319
.223
.193
.140
.358
1978-79
1980-81
1982-83
Number of observation:
Note swes_ples=
Tab
5.193
.281
.317
.153
.429
.497
.361
.407
2.340
.151
.167
1 if hving in an SMSA
1 if year is
1968-69
1970-71
1972-73 ,
1974-75
1976-77
Y6869
.464
,493
.040
TIMEWORKING
INVOLUNTARY
UN1ON
P.4RTTIhfE
.314
Std.
Dev.
.6il
cOUege graduate
16+
Afean
1 ifmamied
I ifmarries during interval
high school graduate
college dropout
SCHOOL=12
SCHOOL= 13-15
LATE
Std.
Dev.
.3i3
—
““
Table A3: Sample
(SAMPLE
“Early”
Means and Stmdard
and “Late”
Cohorts
Deviations
of 24-31 Year Old Lien)
EARLY
Variable
NoNm7HITE
MARRIED
WEDS
DIVORCES
SCHOOL= 13-15
SCH00L=16+
AGE
PRIOREXP
TIM EWORKING
1NV0LUNTAR%
TENURE”
.195
.396
.274
.446
1 ifdivorces
during interval
1 if child is born during, interval
.773
.037
.0.21
.177
.419
.190
.142
.651
.058
.030
.172
.477
.235
.172
high school graduate
college dropout
co~ege graduate
.379
.156
.238
Yeas
involuntarily
of tenure
Log of red hourly wage
1 if works less than 35 hours per week
1 if wages set by union
I if industry is
Construction
Trasp.,
communi~tion,
utilities
CONSTRUCTION
Trade
Public administr~ion
Unemployment
rate
1 if fivingin the SOuth
1 if fivinginan
SMSA
1 ifyearis
1966-67
1968-69
1970-71
1972-73
1974-75
1976-77
19.78-79
1980-81
TRADE
PUBADMIN
UNEkIPLOYMENT
SOUTH
SMSA
Y6667
Y6869
Y7071
Y72?3
Y7475
Y7677
Y7879
Number
1 if nonwhite
I if left last job
TRANSPORTATION
Y8081
Number
Mean
Yems of potential prior experience
experience
~atio of actual to potential
WAGE
PARTTIkIE
UNION
of individuals
of observation!
Note: smes_ples=
Tab
Std.
Dev.
Definition
1 ifmamied
1 if marries during interd
BIRTH
SCHOOL=12
LATE
Std.
Dev.
50
.363
..426
.394
.235
.219
.424
..414
2201
26.644
4.196
3.279
4.9~5
2.074
3.384
.306
.336
.467
.302
.033
.178
.0.53
.223
3.295
2.822
3.555
2.844
2.125
.457
2.046
.429
.028
.153
.165
.360
.028
.273
.165
.445
.085
.081
.136
.062
.279
.274
.101
.106
.301
.308
.343
.242
.154
.075
3.333
.350
.709
1.309
.477
.454
5.694
.438
.717
.361
.263
1.260
.496
.450
.077
.269
.287
.242
.126
—
—
.266
—
.444
.452
.428
.332
—
—
—
—
—
.147
.299
.288
——
.265
1,438
28,220
.
.377
.489
27.248
853
20,019
1.
.381
.485
Mean
—
—
.354
.458
.453
.442
~~
-
Table A4: Smple
(SAMPLE:
Voluntary
Transitions,
.
Means and Stmdard
Job-to-Job
Deviations
TrmsitiOns,
md
for Women
Continuously
Voluntary
Mean
1 if nonwhite
1 if married
1 ifmmries
during interval?
Mmried
Weds
1 if di~,orces during interval”
1 if child is born during intert~”
D1vOrces
Birth
Newborn
Pr-chool
School
Age
PriOrexp
Timeu,nrking
Involuntary
Tenure
Parttime
Wage
1 if child ageone or under
1 if child age six or under
Years of sclloobng
{ears of potential
Ratio ofactud
topotentialexp.
1 if left I=t j,ubinvol,untarily
Years of tenure
.
if works less than 35 hOrrrs/week
Log of red hourly wage*
1 ifwrages set by union
1 if industry is
Union
Construction
Transportation
Trde
prior ~perience
Const., a~iculture,
mining
Trarrsp., communication,
utifities
Trade
Public adrnirristratiorr
Unemployment
rate’
1 iflivingin
the SOuth
I if living in an SMSA
1 if ye= is
1968-69
1970-il
19i2-73
1974-75
Pubtitin
Unemployment
SOutb
SMSA
Y6869
Y7071
Y7273
Y7475
Y7879
Y8081
Y82S3
Number
of indi
Number
of observations
.434
.498
.270
.248
.489
.068
.195
.031
.095
.015
.044
.188
.049
10.090
.397
.772
.319
.200
.417
2.168
3.901
3.439
.337
.271
1.890
.383
.448
.345
Iuds
.045
.200
12.61 i
24:247
2.858
.403
.0i8
1.913
.126
1.639
.143
.122
.028
.206
.391
.216
3.083
.489
.024
.196
.075
.142
.234
.193
.116
.126
.078
.033
1105
1976-7i
1978-79
1980-81
1982-83
Y767i
Mean
.251
,540
.079
.040
.115
.042
.224
12.945
24.526
3.078
.391
.080
1.647
.178
1.628
.138
Tmnsitions
Std.
Dev.
11,394
.420
.263
.349
.424
.394
.320
.332
.269
.178
‘
.065
10.380
.329
.777
.113
.209
.167
.13i
.162
.084
.094
Workers)
Continuously
Job-t-job
Transitions
Definition
Employed
Std.
De\r.
Employed
Std.
Mean
.432
.261
.500
.252
.172
.294
.207
.400
2.052
4.064
3.262
.347
.268
2.166
.332
.453
.350
.488
.061
.028
.088
.026
.172
13.503
27.140
2.544
.414
.0i3
3.918
.076
1.865
.216
Dev.
.439
.239
.166
.283
.160
.378
2~8~
4.i3i
3.402
.391
.261
3.6s5
.264
.394
.412
:016.
.126
.397
.246
.083
.098
.081
.277
.29i
3.184
.470
.417
11.118
.386
.792
.316
..407
.3i3
.343
.368
.278
.292
.092
.122
.128
.137
.147
.165
.154
.183
.035
653
3,559
.052
.135
.272
3.327
.48i
.406
.2?2
.289
.32i
.334
.343
.354
.341
.187
.390
789
.30,570
- Since we only know that a change occurs between successive interviews, the variable equals one ior every thre-montil
intervalftifing betw=n the two interview dat-.
b In 1982 doUars.
‘ Unemployment rati during first month of tbe interval.
51
Table ..45: Smple
(SAMPLE:
Voluntary
TrasitiOns,
Means and Standard
Job-to-Job
Trmsitions,
Deviations
ad
for Men
Continuously
Employed
Workers)
.-
VOluntaryTransitions
JOb-t@jOb
Transitions
Std.
Vuiable
Nonwhite
Mamied
Definition
1 if nonwhite
1 if married
M=n
Weds
Divorces
1 if m=ries during interval”
1 if divorws during intervfla
1 if ctid is born dufing interval”
.070
.017
.158
13.116
25.264
3.587
.483
.093
1.992
.069
1.987
.117
Birth
Sdool
Age
PriOrexp
T!meworking
Involuntuy
Tenure
Pxttime
Wage
Union
.169
.603
Years of schoofing
Years of potential
prior
experience
of actual to potential exp.
1 if ldt lmt job involuntarily
Ratio
Years of tenure
1 if ’works less tha 35 hOur~/week
Log of real hourly wageb
1 if wages set by union
1 if industry is
Construction
Transportation
COnst., a~iculttiie,
milling
Transp., communi~timr,
utifities
Trade
Pubadmin
TJnemplOyment
Trade
Public administration
Unemployment
rate’.
South
SMSA
1 if Uving in the South
1 if fiving in an SMSA
1 if ye= is
Y6869
Y7071
Y72?3
Y7475
Y7677
Y7879
Y8081
Number
1968-69
1970-71
1972-73
1974-75
1976-77
1978-79
1980-81
of indi
.108
.057
.209
.046
9.620
.402
Dev.
.3?5
.489
.255
.129
Mean
.233
.540
.511
.322
.014
.144
12.534
23.776
.2.925
.478
.111
1.564
.081
1.922
.082
.365
3.034
4.003
3.340
.346
.290
2.230
.254
.311
.141
.233
.043
.407
.189
.210
.050
2.937
.490
8.739
.434
.704
.457
.706
.109
.189
.311
.108
.172
.220
.140
.310
.377
.415
.347
.209
.276
.087
.127
.133
.031
.173
.391
.068
.037
Std.
Dev.
.423
.498
.249
.117
.351
2.930
4.147
3.091
.362
.314
2.097
.273
.546
‘:2-75
.348
.204
.392.
.218
3.107
.496
.456
.406
.447
.281
.333
.340
.252
.188
Continuously
Employed
Mea
Std.
De>,.
.200
.400
.689
.060
.463
.237
.142
.021
.174
13.089
26.222
3.022
.497
.060
3.316
.038
2.081
.176
.084,
.076
.158
.059
9.571
.431
.711
.109
.148
.134
.152
.150
.140
.127
bctwen the two interview dates.
b In 198? do~ars.
= Unemployment rate during first month of the interval.
52
-
.379
2.938
4.571
3.463
.390
.237
3.412
.192
-
.4i7
.3$1
.278
.265
-
.364
.236
3.010
.495
.454
.312
.355
.341
.359
.357
.34i
.333
iuds
1066
935
1353
Number of observations
5,615
49,406
11,841
- Since we only know that a cl, ange occur. between succesi,.e interviews, the vmiable equals one for every thre-month
interval fd~ng
“
“
‘