me Nauoti hngimdtid
Suwew (NLs) progrm sup~rts mmy studies &signed
m iw~e
mdmstmdtig of tie labor m~ket md metiods of data mllection. ~e Discmbn
Papers sti~
%m~rat~
some of he resemh comkg out of the ~ogrm.
Mmy of tie paws in tis smi.s have been dcfivmed x find reprls co-issio”d
~
Pt of m exuaurd
sat
progrm, udm which grmti xc awmded tiou&
a
competitive POC-S. ~e s=ies is titmded [o circtia[e tie findings to htereswd reti~
witi ad outS& tie BUUU of Lahr S [atis[ics.
Pasms titaeskd in obtatimg a wpy of my Discussion Paper, othw NLS re~rs, or
gmerd Smatio”
shut tie ~
pogrm md suweys should co”lact he B“rem of
bhr
St~ktics. Office of Emnomic Rese=ch, Wmhhgto”, DC 20212-OW1”QOZT60%T
7405),
Material h tis publiwion is m the public domtin md, wik appropriate mtit, may
& rqrducd
witio”t patission.
.
.
@ti”om ad conclusions expr.ssd in tis document me tiose of tie autior(s)
md do not necwsztiy rqresent m official posi[im of tie Bur=u of hhr Statistics
or the U.S. Dep=timr of hbor.
,
.
NLS
Discussion
Papers
4
.
<
●
✎
Patiicipation in Low-Wage Labor
Markets by Young Men
Battelle
R. Mak Gntz
Human Affairs Research
Thomas
Stanford
Centers
MaCurdy
University
September
1992
.
frmded by the U.S. Department of Labor, Bureau of Labor Statistics
under grant number E-9-J-9-W46. Opinions stated in this document do not necesstiy
represent official position or pohcy of the U.S. Department of Labor. We gratefully
achowledge many useful comments and expert research assistrmce from We &agg,
Wei-Yln Hu, and Jason Scow
~is
project
was
Exeeutive
Some of the most controversial
Summary
issues encountered
govemmenti
titervention
in bbor
markets involves
circumstances
underlying
individuals
participation
the kbor
market
suggest different
tidividurds
become
contribute
anti-poverty
to household
program.
markets as temporary
experience
positions
minimum-wage
~, wages in such legishtion.,
jobs in economic
mobihty
offering
concerning
in low-wage jobs.
Nowhere
Those holding
policym&ers
individuals
Competing
horn these jobs
raising the Mum
wage as an
who view participation
the oppo~tity
in low-wage
to acqutie
the work
jobs frequently
or for the @elusion of sub-minimum
These two ,divergent
views of
is MS more evident
skills needed to move on to higher-paying
legislation
the
tie view rhat some
jobs and that the -rigs
income advocate
Akernatively,
and tie additiond
argue agtist
legislation.
trapped in low-paying
significantly
questions
pohcy prescriptions.
tian h the case of -urn-wage
in pubhc debates over
training
views of, ~e role played by low-wage
are often at tie root of many disputes
over labor-market
poficy .
This report develops
low-wage
earqings
a comprehensive
picture of the experiences
sector and the role that this experience
prospects.
This pictire
career paths of. individuals
identifies
plays h future employment
where low-paying
employment
in the @itid stages, of their fife-cycle,
ES- 1
of youths in the
and”
fits “into the
and how these career
paths vary across mce-ethrdc
our empirid
analysis
groups and education
attainment
using data from the Natiorrrd Longitudind
~V.
YOUtiS are, of course, an ided demographic
low-wage
sectors because they are major participmts,
of their fifetime employment
experiences
provided
for men.
activities.
Survey Youth cohort
group for stidying
involvement
and they are in the formative
The availability
by the NLSY on a diverse
We carry out
of the rich information
group of individuals,
in
years
on job
in conjunction
.
wifi
.
long obse~ation
Priods,
term consequences
such employment
offers the opportunity
of involvement
in low-paying
involvement
framework
based on a sophisticated
synthesizes
the work experiences
This framework
that individuals
tiese activities:
kbor
markets,
including:
employment
appears to in-fluence subsequent
To characterize
career:
to assess both the short- arrd tie long-
describes
employment,
modification
of a transition
Though
our p@ary
our empirid
employment
dufig
the distribution
spend in various activities
model describes
in low-wage jobs,
we formulate
probabdity
the initial ymrs
an empirical
model that
of &eir work
of the lengths of tie number
and the paths that hey
interest
whetier
mobfity.
in low-wage
of individds
and to determine
is to characterize
,.
.,,
foHow to and from
participation
in low-wage
a wide variety of labor-market
employment
of weeks
in high-wage
jobs,
activities
simultieous
.
employment
in low- and high-wage jobs, involvement
nonemployment.
Our empirical
work conducts
@igh-schooI. dropouts,
high-school graduates,
and co~ege graduates),
and for 3 race-ethnic
in educational
separate analyses
individuals
and
for 4 education
Blacks,
.—.
groups
witi some college education,
groups whites,
ES-2
activities,
and Hispanics).
.
-
Fufier,
we extine
the irnpli=tions
one given by the lowest quintile
Current
Population
minimum
hbor
of the wage distribution
of a low hourly wage rate:
for dl workers
in the montiy
and a second equal to $1 above the fededy
legislated
wage.
Our resulting
*
Survey;
of 2 distinct definitions
empirical
markets and of earnings
model captures
6 aspects of participation
in low-wage
mobility:
.
●
The probabdity that a youti
portions of his early career.
wiu enter low-paying
●
The likeMood of entering or returning
alternative labor-market statuses.
employment
to low-wage
employment
The lengths of spe~s spent in low-wage employment,
in high-wage jobs, in nonemployment,
in edumtiond
combinations of these statuses.
as well as in employment
activities, and in
The Wefihood fiat a low-paying employment speU wfil end with a Participant.
moving to high-wage employment, educational activities, of nonemployment.
●
Measures of the total time spent in low-wage employment--or
in any other state-over an ex~nded period of-time, includin~ the ;um-ber of spells an-d tie
cumulative duration during a specified. horizon.
To summarize
manner,
the implications
of this empirical
our analysis implements
a simulation
during the fwst 10 ymrs”””after Ieaving.forrnal
.
from
●
@ The way in which experiences i~
among
characteristics and labor market histories.
.
over various
individtis
possessing
different
model in a readily understandable
approach
schookg
to portray
labor-mmket
mobility
in a variety of scenatios.
.
The empirid
evaluafig
findings
of this study offer a valuable
many of. the assumptions
maintained
ES-3
source of information
by policy mtiers
~ heir
for
deliberations
over hbor-market
to answer
●
poficy.
tie foflowing
The report undertakes
Among young men, who participates
. Is employment
Disputes
Summarv
employment
a porf-of-en~
sectors?
iobs contribute
to income?
in low-wage’ markets?
sectors temporary
to these questions
interventions
or ~rsistent?
into high-paying
iobs?
are often at the heart of many pubhc debates
in the labor mWket.
of Bas ic Findings
A synthesis
●
from low-wage
in low-wage
over tie answers
over govemmenti
in low-wage
less stable for workers
IS participation
. Is low-wage
of these results in an effort
questions:
. How much does earnings
●
a synthesis
Among
of our findings provides
Young
Men,
the following
Who Participates
answers
in Low-Wage
to tie above questions:
Swtors?
AU demographic groups have some ex~rience.
High-schooI
drdpouts ‘mdhigk-school
gradua~s have a 75 %-80% chanceof holding a
low-wage iob at sometime during tie fwst 10 years after Ieaving school.
The participation rate of co~ege graduates ti the low-wage secfor dutig
their first 5 years after school is a surprisingly
high 30%.
.
High-school graduates experience tie mosf time in low-wage jobs
during the 10-year period after school, followed by high-s~ool
dropouts,
with college-educafed
individuals having considerably less experience.
The
larger amount of low-wage employment for high-school graduates rebfive
to dropouts reflects that a significant segment of the dropouf population
spends a considerable amount of time h “ionemployment.
High-school
dropouts and graduates average 1-2 years of low-wage employment during
the 10-year period. Approxima@ly 50% of high-school dropouts start their
ES-4
employment in low-wage jobs; this ~rcenmge
steadily drops for higher
levels of educahon, fting
to well below 20% for college graduates.
Measurements
designed to gauge the extent of low-wage employment
depend critidy
on the dme frame used to register pardcipation,
as well as
on the age of individuals.
Consider the popufatiun of high-school dropouts
and high-school graduates.
If involvement h low-wage labor snarke~
refers to holding any low-paying job during a 2-year period when tiese
individuals are in their late teens, tien the participation mte in low-wage
employment is as large as 59%. Alternatively,
if low-wage experience
means holding a low-paying job during any particular quarter when these
individuals are in their krte twenties, then the participation rate is as sma~
ash%.
.
.
.
Regardless of tie age and education of young men, participation ti
low-wage jobs is twice as Uiely to occur over 2-year horizons tian in any
pardcular quarter making up his horizon. respective
of tie time frame
used to @talog participation,
tie IAelihood of holding low-wage lobs
decfines with age @y as much as 50%), with the most signifiwt
decreases
occufig
at the earfier ages. As individuals age, education becomes a
stronger factor determining the likelihood of holding a Iow-w?g.e job.
Race-etinic
differences in low-wage participation are relatively
minor at young ages, but they tend to widen at older ages especidy
at
lower levels of education.
Ignoring coUege graduates, Blacks tend to have
the highest participation rates in the low-wage sector, with no systematic
ordering between Whites and Hispanics.
Race-ethnic differentials are
inconsequential
for college graduates.
.,,
.
●
. ..
. . ... . .. .. . . . . .
How Much
..
Does Earnings
from
Low-Wage
Jobs Contribute
to””hcome?
The fraction made up by earnings from low-wage jobs falls as one
considers longer horizons for accumulating income and broader sources of
income. Earnings from low-wage jobs is a relatively high fraction of
individuals’ labor income earned during the quarter during which these jobs
were held; it is a much smaller fraction of toti fatily income received
over 2 years. For high-school dropouts and high-school graduates, earnings
from low-wage jobs average as much as 86% of totil labor. income during
those quarters during which such jobs were held; this average drops to as
btie as 30% when calculating the contribution of these earnings to totil
ES-5
fdy
income received over 2-ywr periods
employment occurs. For college graduates,
71% and 15%.
during which low-paying
the analogous percentages
are
For less+ducated
men who hold low-wage jobs, tigs
from these
jobs account on average for 65%-80% of t.oti family hcorne received in a
quarter, and 40%-50% of ftiy
income received over 2-year periods when
low-paying jobs are held. Considering individuds’ labor income alone, the
fraction of -rigs
coming from low-wage jobs for IessWucati
pardcipants ages
between 69%-88% when considering the contribution of
these tigs
to quarterly labor hcome; and it ranges between 33%-61%
when considering the contribution to toti labor income received during 2ya
periods when low-wage jobs are held.
The contributions of earnings from low-wage jobs to any measure of
income over any period declhes with age and with higher levels of
edumtion.
The rate of decke
associated tith aging is much steeper for
higher education levels. At the lower levels, the fraction of any m~sure of
income made up by earnings from low-wage jobs is tyljdy
highest for
Blacks, with Hispanics second, and with Whites having the lowest fractions.
Race-ethnic differentials are insignifimnt for college graduates.
●
k Employment
MS
Stable
for Workers
in Low-Wage
Labor
Markets?
A centd finding in MS study concerns the role of work experience
The tod amount of past employ-merit is
in future labor-market activities.
the main factor predicting the extent of future employment, irrespective of
whetier Mis past experience occutied recentiy or whetier it occumed in
low- or ~gh-wage jobs. ““Withrn’ariy =ducation group, “past em”ployment is
tie primary factor governing the stabi~ty of individual:’ future
employment.
The N
between previous labor-market ex~rience and
future employment diminishes at higher education levels. A hge
component of tie race-etitic
differentials observed in the amount of
employment and tie fraction of this employment spent in high-wage jobs
reflects the influence of hmited. employment early in mreers--i. e., time
spent out-of-work, not time spent in low-wage jobs--that has a compounding
effect on subsequent labor-market activities.
●
E Participation
in Low-Wage
Sectors
Es-6
Temporary
or Persistent?
.
.
.
There is a great deal of mobfiity out of low-wage jobs. The
evidence d~s not support tie notion of a low-wage “trap”. Young men
spend rehtively fitde time in low-paying employment in tie fust 10 years
after school. hdividuds
typidy
experienm very few entries into lowwage jobs during this period. Even for the lowest education groups, only
50% start low-paying jobs 2 times or more during tie 10-ys~. Pcrid; only
10% start such jobs more than 4 times. Further, the durations of spells in
low-wage employment are surprisingly short. TypiA
SWUS last for leas
tian 5 months for the high-school dropouts, and around 6 months for highschool graduates and beyond; 90% of M speUs are compIeted weU withti 2
y-s.
Wgh-school dropouts and graduates average a modest 1-2 years of
low-wage employment in total during the 10 years after school; the majority
spend less tim 1 year; and ordy 10% of this population spend more than 4
yms in low-paying jobs.
About 50% of high-school dropouts or graduates centering low-paying
employment find a high-wage job within 15 months; 90% fmd such
employment withii 3-5 years. Whale tie typid
members of race-ethnic
groups wait S*
amounts of time to enter high-paytig
employment, a
component of Black high-school dropouts experience exceptionally long
waits; 25% are not in high-paying employment before 3.5 years; and 10%
are not in such employment even after 9 years. Most of &is wait,
however, does not reflect long periods spent in low-wage employment,
instead it reflects much hme spnt out of work.
Mobdi~ out of low-wage jobs does not vary systematim~y across
education levels, except for college graduates who experience noticeably
more rapid movement into high-wage jobs. Substantial differences exist in
mobfi~
out of low-wage employment across race-ethnic groups at lower
Ievek of education, but these differentials steadily dissipate for the higher
education groups and become inconsequent
for college graduates.
.
If there is any notion of a “trap” in tie labor market, it must
hcorporate
time spent out-of-work ~ong with low-paying employment.
Time spent ti tie combined state of nonemployment
and low-wage jobs
provides a notion of a Iab.or-market status more relevant than time spent in
low-paying employment done for judging tie prospects of individuals in the
subsequent stages of their working careers.
The extent of phcipation
in
tie combined status of low-wage jobs and out-of-work d~s.hes
sharply
for higher levels of education and for older workers.
ks-iducated
Blacks spend an exceptionally
ES-7
large amount of time in
tis status, and the experiences for some poorlydumti
Whhe and
Wsparric dropoufi are not much better.
More than 50% of Bkck highschool dropouts spend more than 6.5 years in low-paying jobs or out of
work dutig
the first 10 years after school; and 10% spend more. than 9
years. About 50% of ~spanics
spend about hdf of tie 10-year period in
low-wage employment or out of work; and 10% participate rdmost 9 years
in these statuses.
White dropou~ fare better than Blacks and Wspanics,
and yet 10% of Whites spend 7.5 years or more in low-wage jobs or out of
work.
The hge amount of time spent in this combined state of low-wage
employment and nonemployment
primarily reflects the influence of
insufficient bbor-m=ket
experience early in .rhe career which predicts low
future employment.
When individuals acquire labor-market experience,
even in low-wage jobs, their prospects for future employment are
signifi~fly.
enhanced.
●
h Low-Wage
EmpIoyrnent
a Pott
of Entry
into ~gh-Paying
.
Jobs?
hcreaskg
employment at any wage implies more time spent in highwage jobs in the future. The evidence supports the famdiar concept of fifecycle wage growth, which depicts participation in low-wage sectors as
offering individutis the opportunity to acquire tie work experience and
additiond skills needed to move on tg higher-paying jobs in the future. The
acquisition of work experience in low-paying jobs at earlier ages not only
impfies more employment in the future, but a greater amount of this future
employment takes place in high-paying jobs.
“~ The ftidings gtierally
support tie view that tie majority of people
entefig
high-paying employment will not experience low-wage jobs for
quite some time, though a nontrivti
fraction of the high-school dropouts
and graduati
become involved in low-wage employment in the not-toodistant fiture.
From drne of entry into a high-wage job, 25% of highschool dropouts are in low-paying,jobs
~thin about a year. On” the other
hand, over 25% avdd..such jobs for over 8 years.
factor
wage
wage
about
Once in a high-wage job, edu~tional attainment is a signifi~t
in determining the length of time before participation occurs h lowemployment.
Whereas 75% of tie coUege graduates who srart highjobs wait at least 5 years before entering low-wage employment,
50% of high-school dropouts are back in low-wage jobs within 3.5
ES-8
.
years. Ram+thnic
groups have broadly similar expenen=s
ti entering
low-paybg employment after worting at a high-wage job, regardless of the
levei of ducatiomd
atiment.
,
.
.,
. .... ...-... .,,’
, .,
...:-..:
.... . . .. .. .... . .. . . ,., .
.
.
ES-9
... . . .
. .
.
.
.... . . .
..
CONTENTS
Page
Section
Executive
.: ES-I
Summary .. . ..... . .. . .. . .. . . . . . . . . .
1-1
1-6.
1-9
1
Mhoduction . . . . . . . . .. . . . . . .
1.1
Overview of the Literature . . . . . . . .. . . . . .
1.2
Outine of the Paper . . . . . . . . . . . . . . . . .
2
Data Description= .. .. . .. . ... . . . . . . .
2.1
Sample Selection Criteria . . . . . . . . . . . . .
2.2
Defintig Economic Statuses . . . . . . . . . . . . .
2.3
Descriptive Statistics . . . . . . . . . . . . . . .
Descriptions of Durations . . . . . . . . . . . . . .
2.4
Descriptions of Transitions . . . . . . . . . . . ...
2.5
3
Earnings from Low-Wage Employment . . . . . . . . . . . .. . .
Ernpiricd F*ework
. . . . . ... . . . . . . . . . .
3.1
3..L
Participation in Low-Wage Employment ... . . . . . . . . .
Relative Importance of =mings
from Low-.Wage
3.3
Jobs . .. . .. . . . . . . . .. .. ...
3-L
3-2.
3:3
An Empirical Ftimework . . . . . . . . . . . . . . . . . . . . . . . . .. .
Probabilities Determining Initi4 hbo.r-M~ket
4.1
S“tatus. . . . . . . . .. .. . . . . . .
Duration Distributions and Hazard Rates
4.2
Eqti_~
Probabdities . . . . . . . . . . . . . . . . . . . . . .
,4.3,,
‘“’”4.4’: CharactirifingE mpIoymentExperiefis.....:....:
4-1
4
.
5
Empirical
5.1
5.2
5.3
5.4
5.5
5.6
6
-.
Specificahoris-tid
Esti”mation . . . . . . . . . . . . . . . . .
Parameteritition
of Dumlon
Distributions and
Ha=d
Rates . . . . . . . . . . . . . . . . . . . . . .
Estimation of. Duration. Distributions
and Hazard
Rates.., . . .. . . . . . . . . . . . . ..... ... . . . . . ..
Parametenzations
of Entrance Probabilities . ...Estimation of Entrance Probabilities . . . . . . . . . .
Parameterimtions
of Initial-Status Probabfihes...
“Estifiation of Inihal-Stams Probabilities.” . . . . . . .
Characterizing
Labor Market Experiences and Mobifi~ .. . . . . . . .
Simulating Employment Experiences.....:..
6.1
6.2
Cumulative Experiences . . . . . . . . . . . .
2-1
2-1
._2-3
._ 2-5
2-6
..217
3-7
44
. 44
4-6
:- 4-7
5-1
5-1
.=..5-5
.5-7
5-8
5-10”
5-1o
6-1
. 6-3
6-4.
Low-Wa~e Em~Ioyment and Nonemployment..
High-Wa~e Em~l~yment and Traini~g..:
Differences in Findings Based on LQ and M
De finitions. . . . . . . .
6,2.4 Summary of Findings for Cumulative
Experiences. . . . . . .
Numbers of Episodes . . . . . . . . . . .
6.3
Lengths of Episodes . . . . . . . . . . .
6.4
.Durations
To and From Participation in Low-Wage
6.5
Jobs . . . . . . . . . . . . . . . . .. .
6.5.1 Durations Urrti First Jobs . . . . .
6.5,2 Durations To and From Low-Paying
Employment After First Jobs . . . .
6.5.3 Time Between Low-Wage and Training
Experiences . . . .
6.5.4 Summary of Findings for Durations To
and From Low-Wage Jobs . . . . .
6.G” ““--Liting
Work Histories and Future bbor-Market
Activities . . . . . . . . . . . . .
6.6.1 Relationship Between Previous Work
Experience and Future Employment . . . .
6.6.2 Effects of Training and Composition
of Past Employment on Future Work
Activities . . . . . . .
6.6.3 Summary of Findings for Work Histories . ..
62.1
6.2.2
6.2.3.
.
7
Conclusion
7.1
““:””’””” 7.2
7.3
.
7,4
7.5
7.6
6-6
6-10
6-12
6-13
6-15
6-17
6-21
.6-22
6-25
6-28
6-30
6-31
6-34
“6-35
6-36
7-1
Among Young Men”, Who Participates- tin’
Low-fiage S~ctors? ~.......
How Much Does Earnings From Low-Wage
Jobs .Contribute to Income? . . . . .
Is Employment Less Stable foi Workers
h Low-Wage hbor Markets? . . . . .
Is Pardcipation ii Low-Wage Sectors
Temporary or Persistent? . . . .
Is Low-Wage Employment a Port of Entry
into Mgh-Paying Jobs?- ~....
Concluding Remarks . . . . .
7-2
.:..
7-3
7-4
7-5
7-7
7-8
References
A
Appendti
A.1
A.2
A
The Sample . . . .
The Variables . . .
A-1
A-1
A-2
Experiences . .
A.2. 1 hbor-Market
A.2.2 Regular School A&ndance . . . . .
A.2.3 Pardcipation
in Otier Edu=tiond
Programs . . . . . . . . . . . . . .
A.2.4 Sources and Amount of Nonlabor
Income . . . . . .
A.2.5 Demographic
Variables . . . .
B
Appendix
B. 1
B.2
B
An Econometric Framework for Estimating Duration
Distributions . . . .
Empirical Specifications of Initial Status and
En-cc
Probabilities . . . ..
,-..,.:..
.
—.,
,..
A-3
A-5
A-7
A-7
A-9
B-1
B-1
B-12
.,,,.—
.
FIGURES
AND
TABLES
Figures
1
Low Wage Defihions
Tables
2.1
2.2-LQ
2.2-M
2.3-LQ2.3-M
2.4-LQ
2.4-M
3.1-LQ
3.1-M
3 .2-LQ
3 .2-M
6.1-LQ
6.1-M
6.2-LQ
6.2-M
6.3-LQ
~ .6:2-M.
6.4-LQ
6.4-M
6S-LQ
6.5~M
6.6-1
6.6-2
6.6-3.
6.6-4
Summary Statistics of Individud Labor Market Experiences by
Education Category
Summary Statistics for Spells by Education Group
Summary Statistics for SpeHs by Education Group
Summary Statistics for Entices
by Education Group
Summary Statistics for Entrances by Education Group
Summary Statistics for Exits by Education Group
Summary Statistics for Exits by Edutition Group
Participation Rates in hw-Wage
Labor Markets Percentages for
Young Men by Age and Education
Participation Rates in Low-Wage hbor Markets Percenmges for
Young Men by Age and Eduwtion
Percen@ge of Earnings Received from Low-Wage. Employment
Average Percentage by Age and Edumtion
Percentage of Earnings Received from Low-Wage Employment
Average Percentage. by Age and E@u~tiorr
Summary Sbtistics for Cumulative Experiences
Summary Statistics for Cumulative Experiences
Cumu!a.ti.ve Experiences Duting 10 Years Following School
Cumulative Experiences During 10 Years Following School
Number of SpeUs During 10 Years FoUowing School
Number of Spefls During” 10 YWS Foflowing School
Lengti.of Spells Du.@g 10..Years Following SCEOO!
Length of Spefls During 10 Years Following School
Durations” To And From, Low-Wage Employment During 10-Year
Horizon
Durations To And From Low-Wage Employment During 10-Year
Horizon
Cumulative Weeks During Three Years Follotig
Particular TwoYear Work Experiences for High-SchooI Dropouts(l 1-)
“Cumulative Weeks During Three Years FoUowing Particular TwoYear Work Experiences for Mgh-School Graduates
Cumulative Weeks During Three Years Following Particular TwoYear Work Experiences for Some” College(13- 15)
Cumtilative Weeks During Three Years Following Particular TwoYear Work Experiences for College Graduates(l 6+)
AC~OWLEDGEMENTS
Ttis project was funded by the U.S. Department of Labor Bureau of Labor
S@tistics under Gmt Number E-9-J-9-M46.
~Inions
s~ted ti tiis document do not
necestily
represent the officti position or policy of the U.S. Department of Labor. We
gratefu~y achowldge
many useful comments and expert research assis~nce from Mke
Cragg, Wei-Yin Hu, and Jason Scott.
. ..”.
:.,
. . . ,.
.,..
:.
,.
1. btroduction
Charactetiing
kbor
markets
tie circumstances
invariably
occupies
Two divergent
the formulation
of many public policies.
views of economic
especia~y
minorities
presents
temporary
low-wage
for advancement
with neoclassiml
support contrary
dual. hbor
marke~),
that an individual
is smd.
and rhe additiond
programs
and legislation
minimum
wage legislation,
a
to this view, some
employment.
d~at portrays
A
human
low-paying
jobs as
depicts participation
.,..
the opportunity
in
to acquire the
skills needed to move on to higher paying jobs.
the validity of both these views as the motivation
that intervene
in labor markets.
A frequent argument
for example,. is that employment
and that tie earnings
litera~re
describes
theories. ~f the labor market (e.g.,
mayke~ as a way station offering individuals
in
in low-Paying jobs ~
to high-earnings
mobility
positions
moves far from a
According
Assuming efficiency of markets, this pictire
,-...,
.. . . . .... .. .
.. ... . ..
. . . . ,,
Public poficy presumes
in low-wage
in the economics
become, -pped
a picture of wage and earnings
positions.
,., . . . .
work experience
temporary
distribution
and women,
which tilere are few opportunities
second view, associated
participation
One view, associated
mobility where the like~ood
relative point in the wage or timings
~pital),
mobility
theories of the labor market (e.g.,
world with rninimd
individuals,
hdividuals’
tie center stage of many public debates over labor-
market poficy.
with institutional
.,
underlying
in low-paying
derived from such employment
1-1
constitutes
for
suppordng
jobs is not
a substitial
.,$ ..
component
of family income.
Proponents
of such Iegisbtion
of tie fabor market and argue in favor of increased
progm
that raises the earnings
distribution.
minimum
tie institutional
view
wages as an antipove~
of workers who are stuck at the bottom of the wage
At tie same time, Mose adopting
economic
acmpt
mobifity argue that the imposition
a neoclassic
of constraints
view of the world with high
on wages restrict the
.
employment
in higher paying jobs.
for a sub-mirdmum
wage that defers the imposition
time that a~ows for the training
Numerous
supported
Proponents
programs
provide
training programs
to compensate
for a ~riod
labor markets
is another
mobiliv.
prominent
hinge on the
area in which the views clash.
view declare a need for government-sponsored
for a vari~y
of.{ticm”ral
educationrd
in low-wage
attainments
markets as permanent
of individuals.
of
The desi~biliv
traifing
failures tilat must be overcome
better employment opportunities for some individuals.
These advocates
... .:,
. ... . ..... . .: . ... . . .._-.:
..... .
. .. . . ..... . . . . .. . .. ..
participatiori
of
of workers.
debates over pubhc policy involving
of the institutional
leads to the provision
of wage cmrstraints
of he two competing.
viewsof labor-nlarket
app~cabfliv
pubhcly
The force of such arguments
witiout
intervention
To the extent fiat promoters
to
view
..-”
that increases
the
of tie efficient.
markets
view tend to support training policies at til, they advocate
individuals
to gain entry into labor marke.m-where
employment
involves
after the acquisition
pubfic debate directed
@ey ..can””
progress
of work experience.
towards eradicating
1-2
it as a deviw
beyond
A third prominent
discrimination.
to allow
loW-PaYkg
example
poficy .m~ers
~~
.
an institutional
permption
or otier stictud
of economic
changes
mobifity
plead the mse for preferential
in labor markets to improve
groups to exit low-wage jobs.
earnings
mobitity
Poficymakers
merely advomfi
tie prospects
for disadvantaged
with an efficient-markets
the requirement
programs
perception
for open job oppofinities
of high
for M
groups.
Surprisingly
wage employment.
titde is known about the process by which workers
enter and exit low-
A tige
of poverty,
body of research
this notion” of income variability
This loose tige
describes
is only penpherdy
the persistence
finked to low-paying
in part reflects tie fact that members
of nonpove~
in low-wage jobs, and in other part arises from the inclusion
determination
of poverty.
time series properties
invariably
disperse
There is also a considerable
of earnings
at the individud
analyze either annual earnings
points in rime.
missed udess
.,,
employment
,.
source of earnings.
insufficient
Consequently,
Iriovements
in these markek
households
income in the
body of research
summarizing
the
level, but such studies
or a person’s
wages.. observed
to and from low-wage
is long-te~
work
and represents
at
markets
are
the principal
..,.
Further,
to determine
measures
employment.
of nonwage
or household
but
knowledge
the likelihood
of intertemporal
that an individual
correlation
patterns alone is
fa~s into any particular
earnings
,
category.
research,
Even after integrating
the evidence available
there is far too tittle information
approximating
a complete
wage labor markek,
characterization
from tie various bodies of
to enable one to formulate
of tie types of workers
the extent to which these workers
1-3
anything
who
participate
remain in. such markets,
in 10W-
and the
reasons for exiting Iow-paytig
employment.
This paper develops a comprehensive
picture
of the experiences
of youths
wage labor markets and the role that MS experience
pIays k future employment
prospects,
Survey Youth Cohort
ustig
data from the National
picture identifies
tie tiitial
where low-paying
Longitudinal
employment
group for carrying
attainments
for men.
The avaflabifity
out a study of his sort because
.on a diverse group of individuals,
involvement
tiiti~
in
in low-paying
employment
on job experiences
in conjunction
and to determine
low or high wages.
that is csdmable
the work experiences
distinguishing
activities.
by the
periods,
consequences
of
whether such employment
employment,
. .,.
,
of individuals
among jobs according
we formulate
during
to whether
avtilable
of iridividuaIs,
recogtitig
they pay
model
that the
for each person in the sample covers onIy subsets
years in the fife-cycle of concern
h this analysis.
1-4
To accomplish
an
the
One cannot infer such a picture without devising a statistid
using hmited data on the activities
period of observation
in low-
mobility.
that synthesizes
stage of their life-cycle,
provided
with long observation
To develop this picture of involvement in low-wage
., :...
=.,.,.-.... ,, ...;
,.
.,. . . ... . . . . . ..
framework
demographic
fifetirne employment
to assess both the short- and the Iong-temr
to influence subsequent
empirical
~ls
of tidividuds
they are major participants
years of heir
of the rich source of information
offers. the opportunity
,, . ._
~~.
Youths are, of course, an idd
wage markets and they are in the formative
appears
md job
stages of their fife cycle, and how these career patis vary across race-ethnic
groups and education
WY
fits into the career patis
in low-
of hose
this task, we develop a
sophisticated
variant of a transition
age, rather &an @endrrr
distribution
activities
.
probability
time, as the relevant
of the lengths of the number
broader
range of activities
various
wages, youths spend considerable
trtilng)
movements
in our statistical
simultaneous
episodes
Mis model,
This requires
Consequently,
low-earnings
in low-and
In defining
we consider 2 distinct thresholds
Current
,,
legislated
employment,
Our statistid
with many featires
the concept
high-eatigs
ting,
TPM that allows for
of a low-wage
sector in
hourly wage rates to rhe low
for dl workers h the
~ge.
dependence,
..-
model provides
incorporated
in the Lterature.
experienced
pursuits
employment,
of a “five-smte”
for assigning
at
Population Survey; and a second equal to $1 above the ftiedy
.
.. .
. . ... . ..
. . .
..
minimum
elsewhere
to being employd
our TPM describes youths’
one given by. the lowest quintile of the wage distribution
montiy
.
our primary
we must consider a
high-eamhgs
the development
ti the various states.
Though
arnou.nts of time either in education
status.
participation
h addition
the
spend h various
labor markets,
model.
events using
This hewo:k..describes
tiat individuals
in low-wage
or in a nonemployment
and nonemployment.
multiple
participation
among five distinct activities:
employment,
duration
summarizing
and tie paths that they follow to and from these activities.
is to characterize
secton
time metric.
of winks
interest
(e.g.,
model CPM),
a flexible alternative
in our specifications
Our formulation.
to a competing-risks
not found in empirical
not only accoun~
for different
it also allows for a broad range of labor-market
prior to current episodes
to influence
1-5
present
durations
model,
applications
forms of
activities
and the Iikefihoods of
futire
vw
exits to other statuses.
~ a WY tit
depends
To summ~e
Fufier,
our formulation
on previous
the implications
petits
duration
work history and on demographic
of WIS model in a mdfiy
a simulation
mobfity
ten years after leaving formal schooling
scenarios.
me
fidings
approach
to produce
from these simulations
provide
to
characteristics.
understandable
our analysis implements
during the fist
dependenw
manner,
a picmre of labor-market
under a v&ety
the basis for answering
of
a wide
-
assortment
kbor
of questions
concerning
the dynamic patterns
markets md the relationship
addition,
the results
demographic
provide
allow-us
groups.
considerable
educational
to determine
information
movements
mobfity.
k
differ across age and
from the simulation
about youths’ mobility
as well as their
with subsequent
how tiese experiences
statistics derived
in low-wage
‘
malysis
across jobs offering different
to and from nonemployment
and
activities.
1.1 Overview
,,.
of the Literature
,
., .,,,,
There is no empirical
proposed
finking this participation
The summary
hourly compensation,
of participation
here.
,
. .. . .
research
Some longimdinal
.
of which we are aware that resembles
data sets covering
a Iengtiy
time hotion
the analysis
provide
.
sufficient
information
to s-tidy the intertempord
weeUy wages observed
this hited
level.
at interview
source of information
Analysis
of ~lngs
movements
in individuals’
times” which occur in one-year
has not been investigatti
mobility at time intervals
1-6
shorter
intervals,
hourly or
but even
except at the most elementary
than a year over a long
period has not been possible
offered
w-MY,
monthly
bemuse,
or quarterly
The most wideIy cited studies
individuals’
.
relative
annual bbor
mobfity.
Social Security
indicate
Schfiler
to conclude
earnings
distribution.
support
comparable
seved
years apart to measure
in the most prominent
categories,
between
that workers
Results
Schiller
on average
the degree “of
study, draws a sample from
measures
in 1971.
the movemenrs
Controtig
Dividing
tie
of individuals
for the effec~
moved about four categories,
of age, the
which leads
are highly mobile across relative positions
in the
from several other studies applying similar techniques
SchiHer’s findings.
Kohen,
Parries and Shea (1975) carry out a
analysis using data from the National
1966. ““Calculating movement
of
males who were 16 to 49 years of age and
1957 and 1971.
tiat individuals
no data set has
on she wages earned by individuals.
in 1957 and who had positive earnings
the categories
findings
of rhe NLSY,
of earnings stabflity compare rhe tings
records of nearly 75,000
sampIe hto 20 earnings
generally
information
income observed
SchMer” (1977),
earned at least $1 ~
among
untiI the avdabihty
Longitudinal
Surveys
commenced
~ family income rather than in eamirigs between
h
the years
.’. ,.
1966 tid
Blacks.
.
1969 for young men, this study finds differential
Taubman
(1 975) examines
sample which conrains information
World War fl and classified
earnings
on middle-aged
above average
1969, tiis study Me the others provides
earntigs
mobifi~
in IQ.
mobti~
between
using data from tie ~ER-TH
men who were military
Comparing
etiings
over long periods sta_fiing from most posi.tio.ns”in tie earnings
enlislees
in
in 1955 and
evidence that there is substantial
1-7
Whites and
mobili~
distribution.
in
~
This body of re=ch
hbor
markets.
at best onIy hints at tie story of participation
Two major shortcomings
lwd to this situation.
ady=s
mobfity
ordy in terms of movemenw
income,
and low annual earnings are not synonymous
On the one hand, the annual earnings
worked
armud
during the year rather tian employment
earnings
Fwst, this re=ch
in annual measures
may be high even if employment
of to~l labor or famdy
with employment
of an individud
Second,
at a low wage rate.
the period
years apart.
Consequendy,
bounded by these points is totily
about participation
On the other hand,
sequence
of annual earnings or of annual average-houriy
of years using panel data, which indirectly
tie nature of earnings
Lflard
mobifi~.
Examples
and Willis (1978), Ldlard (1983),
(1983) and MaCurdy
(1982).
above h that it considers
from the fust shortcoming.
difficulties
in drawing
only between
any mobdir~ occurring
the time-series
correlation
earnings over a long successive
provide
some evidence
concerning
. .’
of this line of research include *e work of
Chamberlain
(1984), Bhar==va and Sargan
over a continuous
The use of annual measures
inferences
within
nothing is known
While this work a.v.oids the second shortcoming
info~ation
job
during. tis. period.
There are a variety of”other studies fiat examine
properties
in a low-paying
consider mobdity
missed and absolutely
in low-wage employment
-——
in a low-wage job occurs because hours
studies making up tilis body of research
two points in time set seved
in low-wage jobs.
may be low due to few hours
of work are high, other jobs are dso held, or the time of employment
is short.
h low-wage
about pafiicipation
1-8
noted
sequ=rice .of years, it does suffer
of earnings leads to serious
kr low-wage
markets.
Even
—
. ...
absmcting
from Wls problem,
for assesstig
.
sticture
mobifity,
associated
a researcher
of earnings
problems,
this body of work provides
across employment
with the variabili~
and about tie determination
=tegories
findings describe
in annu~
timings.
only
To
also needs to know much more about the distributional
propeties
1.2
results from these smdies are not we~-suited
the extent of earnings mobility because tfre primary
tie autocorrelation
measure
the empirid
of initial conditions.
very fitde information
paying different
Given these
on the topic of mobifity
wage rates.
OutKrre of tie Paper
Addressing
an understanding
factors leading
tie substantive
tiis research
of tie routes followed by individuals
to continued
tie routes of =cape
succinctly,
issues motivating
participation
from these markets.
and its simulation
time that individuals
.,
typica~y
lifetimes and of the Melihoods
requires at a minimum
into low-wage. labor markets,
tie
in these markets and to remm participation,
Our empiricrd
provides both a complete
spend in alternative
‘., ..,.
-.:
model captures
characterimtion
activities
during
tiese
and
features
of”the lengths of
We early part of tieir
that they move among vwious. activities
ii pficular
sequences.
Specifically,
the subsequent
analysis focu=s
on answering
wo findamsntil
sek of
questions:
. How fiportarrt are earnings received from low-wage employment?
Do
workers with low-paying jobs tend to hold more than one job, simultaneously,
or are hey more likely to work part-time? Are low-wage e.mers ~ely to
come from house hold.. where more than one member WOIkS.or where there.
1-9
are other sources of income? How dms tie composition of pticipants
br
low-wage kbor markets vary by such demographic characteristics
as age,
race, and education level?
. How tempo~
is employment in low-wage jobs? How Wely is it that m
individual who works near the minimum wage sometime during a y= w~
be doing so at various times in tie fiture? Are low-wage jobs “traps” of the
sort envisionti by institutional theories, or do tiese jobs serve as ports-ofen~ into job ladders with stable and higher-paying employment in the
fiture? What is the Welihood of moving from low-earnings employment to
high-paying jobs, and vice versa? What is tie geneml nature of wage and
earnings mobility among the young? Are earnings and employment less
stable for workers in low-wage labor markets? What is the role of Iife-cycIe
wage groti
in economic mobility?
While simple summary
requtie
the empirical
statistics
enable us to answer aspects of these questions,
results from tie multi-state
duration
framework
to
we
investigate
most
of tie issues.
The remainder, of this paper consists o~sti
next, oufines
hdividuals’
the data set used in our empirical
experiences
in employment
and in ‘tiaining and schooling.
presents
earnings
4 outlines
_gs
across education
analysis,
Section 2, which fouows
integrating
witi low and high earnings,
As a “first step in answering
a variety of descriptive
participation
sections.
statistics
designed
the essential elements
We above questions,
and family income.
framework
the pararneteriations
of our statistical model; along with a brief discussion
1-10
in nonemployment,
Section 3
along with the degree to which
to total individual
of our economekic
mobility; and Section 5 introduces
on
to measure the extent of low-wage
and race-ethnic.groups,
from Iow-payillg jobs contribute
information
Section
for characterizing
used in the esdrnation
of the empirical
results.
Section 6
..
analyzes
the results from tie simulation
comprehensive
from school.
findings
description
Fmafly,
of our estimated
of labor-market
Section 7 summarizes
have to ~y about tie questions
. .... . . . .,,
mobfity
TPM model to produm
during
the initial years after depardng
the results and attempts
posed above.
, ... .
1-11
. .. ,,
:.,
a
,., .,
......,,
to interpret
what our
2. Dab
Characteriztig
detied
the process of economic
work histories
dso differentiating
information
for individuals,
in the WSY
each week to cons~ct
of participation
activities
arrdysis by providing
time.
This section
subsequent
hourly wage rates.
on a wee~y
basis.
These wee~y
of”young men form the foundation
a mechanism
analysis;
of employment,
employment.
time series of employment
in other labor market activities,
but
We use the
~
.
Appendix
statises,
h addition,
we can
such as formal schooling,
series capturing
of tie. subsequent
to classify men into various
outilnes tie formulation.
empirical
and high-wage
very
on wages and hours of work for each job held during
a complete wee~y
training and nonemployment,
labor-market
not only dating their episodes
between jobs that pay different
identify periods
m“obility in tie labor market requires
between periods of low-wage
avtiable
distinguishing
Dwcription
economic
the
empirical
statuses
over
of our sample of young men used in the
A presents
a detailed description
. . . ... ,,
of our data
set.
2.1
Sample
Selection
The ~SY
mobfity.
3 ,~3
Criteria
is an incomparable
It ticorporates
data source for wing
both a randomly
young men, and a supplement
designed,
out a s~dy of economic
natiomfly-representative
sample of 2,576 Black, Hispanic,
2-1
sample of
or economically
“.
disadvmtaged
non-Hispanic,
The youths tiernselves
tie kbor
~Y
non-Black
were interviewed
market experiences
covering
annually
in 1979.
Information
about
of rhese young men is drawn from the first ten rounds of tie
in this paper utifizes z data set of 2,699 young men drawn from the
representative
Black and Hispanic
sample of the ~SY
men.4
hclusion
and the supplement
in our da~ set requird
every year of the first ten rounds of tie ~Y
school sometime
condition
beginning
tie years 1978 to 1988.
The sndysis
nationa~y
men, all of whom were ages 14 to 21 in 1978.
between January
is necessary
samples
of young
a man to be interviewed
and he must have ‘:permanentiy”
1978 and the 1988 interview
date.
to ensure that we capture a man’s experiences
in
left
This schooling
from the time he
fust en”ter~the labor market.
The empirical
of educatiorrd
attainment
Specifically,
~l-.
12
below differentia~s
individuals
:’ persons with
11 yetis
:
persons
among young men based on tieir level
at the’ time they “pennanentiy”
we categorize
13-15:
16+
analysis
into four educational
of edticatiori
with 12 years of duation
with
16 ytirs
of education
groups:
or less @gh-school
(high-school
persons with 13 to 15 years of education
: persons
leave formal schooling.
dropouts);
graduates);
(some college);
and
or more (coUege graduates),
4 We exclude the resWndents from the wonomically disadvantaged non-Hispanic, non-Black
supplemental sample because the original selection of this supplemerrti wmple sufferd from
seved shortcomings.
See Appen@x A for detils.
2-2
h d
of our empirid
work, we conduct separate analyses
for mch of rbese educadonsd
groups.
The notion of pemranentiy
leaving school differs for high-school
An individti
compared
to the other edumtion
education
groups is d=med
attending
school during tie sample time frame.
from high-school,
perrnanendy
workd
mtegories.
to have permanency
formsf
and obtain a high-school
participating
2.2
in a trainifi~.prograrn
To describe a population’s
levels over time,
schooling,
someone who graduated
in college would not
exclusive
we classify
labor-market
college.
the wmple periti
to have permanently
out of high-schooI.s
while they are engaged
.
b contrast,
a
and does not return to
left school and entered
If these individuals
diploma or a GED in the future,
Defining Economic Status~
.,, .,: .. . .... . .:.
.,. .,., ,,,
mutually
For example,
during
within a year is considered
the labor force at the time he dropped
school
from one of the three higher
enter tie labor market rrnti they stopped attending
schookg
as
left school at the time he is last observed
for one year and Wen enro~ed
young man who drops out of high-school
dropouts
we dassifi
return to
them as
in .Xese edgcationaI
pursui~.
. . .. . . ..
participation
individuals
in labor markets paying different
as occupying
statuses. in any week.
we refer to them as occupying
,~. .
any one of sk exhaustive
wage
and
When. a person is atrending formal
labor market status or s=te “s”.
Upon leaving
5 The yw-long interruption in schooling for the high-school dropout is important because
someone with an interruption shorter than a yar is considerd
to have continuously attended
school.
2-3
formal schookg,
labor market
an itrdividti
occupies one of five statuses:
(status f); employed
activity
activity
(status e); and neither working
of our analysis.
between
low-wage
As tie debate beween
so markedly
tiese two sectors
in some type of training or
nor pardcipating
and high-wage
h an”education
the institutio~
employment
in the subsequent
hourly wage distribution
Survey (CPS) mples
Accordingly,
analysis:
for “all workers
(termed definition
is a critid
and neoclassical
points out, there is not a we~-defied
in tie economy.
the wage threshold
,,
employed
(sta@s n).
The distinction
tiought
in a low-wage
in a high-wage job (status h); simultaneously
in both a low- and a high-wage job (status b); tivolved
edumtiod
employed
demarcation
we examine
feature
schools of
line between
two distinct ”specifimtions
of
one based on the lowest quintile of the
derived from the monthly
Current
Poprdation
L@; and a second based on an absolute
differential
of $1 above the federally
Compting
the results ob~ine~, for tiese wo COgCePk Of low wages enables us to assess
the influence
’,.
. . .
over tie smple
1 graphs the values of tie two tiesl]olds
perid.
in time.
use it as our tireshold
the tireshold;
wage (termed, definition
M)
,.
defining
low-wage
employment
The third fine in the figure, termed the “lowest quinhle
the fitted value from a regression
polynomial
minimum
of different definitions on our findings.
. ... .. . . ..... ........
. ... . .. .. . .. . ....
.... . . .. .. ..’
Figure
represen~
legislated
of the lowest qLlintile on a second-order
This trend tracts the profile of the lowest quintile closely,
in the LQ definition
trend”,
and we
of low wages to avoid ematic movements
tie raw lowest quintile not only jumps around,
24
it remains constant at
of
common
values of the hourly wage rate such as at $4.00 and $5.00.
Figure
1, the M threshold
exmeds
the LQ threshold
it rises more slower than the LQ values resulting
As is ‘apparent in
by about $.75 in tie late 1970s, but
in about a $.75 shortfall
in tie early
1990s.
2.3
Mcriptive
Shtistim
Table 2.1 presenfi
some summary
statistics
sample of young men used in our empirical
for the four education
employment.
MSY
To account
to dculate
Generally,
?C!o:s
composition
of ~e supplementary
these summary
of individuals
separate statistics
of low-wage
of our analysis
samples),
sample from the
we use weighted
estimation
statistics.
with different
levels of edumtional
attainment.
.$e,four
.e~:qti:nslg:::?s$,..:h:
.av:y!:.oft.:?l“.::psym:nt.
a:d*.,,
fraction
of employment
Further,
th~ incidence
decreases
The table reports
for the
the results in Table 2.1 reveal what one would suspect about the labor
market experiences
.Loo~g .
atiysis.
and for both the LQ and M definitions
for tie nonrandom
(e.g., the incorporation
procedures
.,
categories
of the labor market experiences
spent in high-wage jobs increases
of low-wage
employment
tifi
.=.
..
tie leveI of edu=tion.
and the amount of nonemployment
with increases.. in edu=tioh.
There are, however,
two statistics in this table tiat ae quite surprising.
incidenw
of &aining is relatively
however,
that “training”
constant across the four mtegories.
encompasses
many different activities,
2-5.
ranging
Fist,
the
It should be noted,
fro”m rcen~
into
‘
schooI to training while employd.
be of a very different
“remarkably
character
high participation
Training
undertaken
by college graduates
from tiat by high-school
rate in low-wage
dropouts.
jobs for couege gradua~s.
one out of every five young men with 16+ years of edumtion
low-wage
employment
Second,
sometime after leaving
school.
ex~riences
Summary
is Wely to
there is a
.SPecifi~IY,
an episode
of
Statistics in the later
tables WN shed further light on these issues.
2.4
D-criptions
of Durations
statistics
Tables 2.2 presents summa~
post-school
labor-market
h, both low-wage
designated
activities .O.e., low-wage
and high-wage
employment.
procedures
in Tables 2.2 corroborate
of spells.
as the low-wage
spe~s decreases
wifi increases
individuals
from the higher education
these low-wage
The tables
cutoffi and hose .mark~
for cksifying
employment
relationships
low-wage
groups and
between education
spells are longer and episodes
groups.
Furtllernlore,
in tile level of edu=tiorial
mtegories
employment
-.
the expected
For example,
n).
results for the four edu~tion
are shorter for tie higher edt!cation
however,
f, high-wage
wage plus $1 as the threshold
of traking.
employment;
employment
are used to dculak
the measures.
. .. .
tie characteristics
of nonemployment
of spe~s spent in the five
e, and nonemployment
Again, the tibles present separate
The resti
ad
b, training
Table .2.2-LQ use the lowest-quintile
Table 2.2-M utilize the minimum
weighted
on the duration
the length
atminment.
are less Wely to experience
Finally>
low-wage
spells are longer in length than the low-wage
2-6
~
spells of high-school
2.5
Ascriptions
Tables
dropouts.
of Transitions
2.3. present the sample proportions
follow particular
paths when entering
2.3-LQ” contains
the weighted summary
low-wage
employment
minimum
wage criterion
“origh
the five post-school
in Table 2.1.
the analogous
school.
Specifimlly,
Thus, witi
these entrance
S~tUS.
First, with regard to ~le c~nstant incidence
by co~ege graduates
has just left and
in Tables 2.3 help clarify some of the unexpected
findings
of .tiaining across education
with a higher level of education
status as compared
Second, the last row in Table 2.3-LQ
that some of the unexpectedly
high rate of participation
is the result of tieir taking a low-wage
to their
and Table 2.3-M
in low-wage
employment
job immediately
after leaving
more than 27 to 33% of all Iow-w’age spe~s for this education
group are the result of an individual
moving from status s direcUy to statis
2.7
of
by the term
has just entered.
are much more Wely to enter training from an employment
suggesh
Table
using the
groups the columns must add up to one because
measures
counterp~.
activities.
measures
sate from which an individual
groups, tie results in Tables 2.3 suggest that individuals
,,:..,
... ,..,
.
,,. ,.,e, ..
less educated
that individuals
The rows, designated
are Conditioned upon entering_ a particular
The summary
labor-market
refers to the status an individud
each of the four education
the ~elihood
for rhe lowest quintile interpretation
to define a low-wage job.
status{’, refer to the labor-market
probabilities
statistics
and Table 2.3-M presenk
the column designation
reflecting
f.
Rndly,
.
one other interesting
simultaneously
someone
feature of TabIes 2.3 is the routes through which men
hold both low-wage
to move tito shtus
statuses
and high-wage jobs.
In particular,
it is very rare for
b from any sta~s other than the other two employment
(i.e., status ? and status h).
Tables 2.4LQ
Wetihood
and 2.4-M
fiat individuals
present
the weighted
sample propordons
with a given level of edumtion
follow particular
exiting from each of the six labor market activities;
s, f, h, b, e, and n.
designations
represent
minimum
LQ and M associated
wage” defmiti~ns
refers to the labor-market
designation
of low-wage
has occupied
fich
The
the origin status
and just left, and the
enters immediately
upon
on an individual
of the statistics
leaving a particular
in”TabIes 2.4 reveaIs several
s~trss.
interesting
in the exit probabilities.
First, the relatively small amount, of movement
.. ..... . ... . . . .... .. . . . . ... .. . . . . . .. ... . ... ..... .
,,, . .. . .
from low-wage
employment
in the higher education
a low-wage
to high-wage
categories.
and
of the rows in this table must sum to one because
are conditioned
CasuA examination
the
paths when
the lowest quintile
b these tables,
status refers to the state in which an individud
these exit probabilities
deffition
respe~tively.
state which an individual
exiting from the origin state.
patterns
with these ables
measuring
employment
For example,
to a high-wage job occurs
surprising,
the highest probability
for an individual
and sti~ only 34% of all low-wage
is somewhat
directiy
especially
of moving
wi~h some college
from
using the LQ
spells end with the man moving
into a high-
.
wage status.
directiy
Second, as expected,
to high-wage
employment
there is more movement
than movement
2-8
from low-wage
from high-wage
employment
to low-wage
across W
,..
,,.,, ,.
education
categories.
Third,
wage and a high-wage
it is uncommon
job to a status where they are not worhrrg
“kdividuals
with more education
specifidy
high-wage
there is substantial
employment
show tis
are more ~ieIy
employment,
movement
to nonemployment,
jobs as a jumping
from nonemployment
This suggests
at sdI. Fourth,
to move from training to employment,
than are hdividuals
stone to high-wage
suggestion
for men to move from holding boti a low-
with less edueation.
FlnWy,
to low-wage jobs and from low-wage
.
hat it may be difficult to use low-paying
employment.
The findings upcoming
in Section 6
to be misleading.
.,
2.9
. .. .. . . . . . .
.._
,.
3. Earnings
An understidhg
of tie relative
to individuals
and their households
income
fwst set of questioris” in Section
rehtive
importance
tie low-wage
time hey
sector;
dimensions.
Specifically,
...”..
,..
chamcteristics
variation
mtegories.
is needed to address
the issues raised by the
determinants
hold high-paying
contribution
and tie time hotion
of the
involvement
in
jobs at tie we
to household
income
over which one measures
employment.
analysis of these four factors not only includes a description
by these measures.
groups,
but dso
This analysis
involves an examination
considers
of
of
both of these
to summatie
,,..
,i
and the importance
the relationship between demographic
,.. ,
of low-wage employment,
it focuses on describing
in these factors across age and racial groups witiin
k addition,
jobs as a source of
the extent of individuals’
employmenfi
employmen~
it explores
time effects that ..me common
educational
of low-wage
Four factors are important
factofs vary across demographic
the time paths fo~owed
Employment
in a low-wage job; the relative
of low-wage
A comprehensive
how tiese
importance
the degree to which individuals
from low-wage
the impomce
1.
of low-wage
are employed
of earnings
.,, ,,
from hw-Wage
each of the four education
tile tine Pa~S. of tie four factors by inco~orating
across age and racial groups
categories
3-1
but differ across the
3.1
Emptild
Frmework
Simple regression
s~tistics
reflecting
mobility.
particular,
offer an attractive
the part played by low-wage
These methods
impofince
toti
methods
of low-paying
approach
employment
provide a way to decompose
jobs into age effects and common
to tie i-th individud
summary
in determining
economic
various measures
let the variable yi, denote the ratio of low-wage
income corresponding
for providing
of the relative
rime components.
earnings
in period t.
h
to some measure
Consider
of
the regression
equation
where the variable
year dummies,
po~ynomial
“ag.eil~ refers to the age of individu~
This equation
and 6s and ys are parameters,
in age; evaluating
the polynomial
i in period t, the varia!les
at a particular
dfit =e
induces a third-order
age gives the average
of y
corresponding
to tiat age over the entire period. Imposing the identification condition &
,.,, .,,,
,:, ,,.
,:.._,,.~,.., .,,,, ,. :,,
.,
...,,
Tk = O tiplies “tiat the yk coefficient represen~ the common deviation experienced by’ W
ages in period k.
The following
analysis
four factors- tilat determine
Specifically,
we consider
dimensions
of low-wage
temporary
or permanent
estimates a variety, of regression
tie re!ative impormnce
two types of dependent
employment
oflow-wage
variables
employment.
3-2
the
employment.
that capture various
and different time hotions
nature of low-wage
models to investigate
to uncover
the
The first type of dependent
“--”
variable
ordy distinguishes
whether an individual
over a given time horizon and thus =ptures
employment.
by pitting
income,
or his fatiy’s
consider
four distinct time horizons
(1) d
periods
dependent
designated
intervals,
designated
We estimate
each category
examine
are incorporated
(2) all 6-monti
(3) au one-year
tie fraction of an individud’s
the priods
in low-paying
years 1978,...,1987,
denoti
jobs.
We
1.. .T in “the subsequent
designatd
intervals during the calenda
htervals,
into the
lY periods;
~d
3m
horizon,
(4) ~1 ~o-ye~
2y perids.
distinct models for each of our four educational
we often estimate separate
categories
models for tie tiree race-ethnic
the time paths of the various dependent
model by including
.,, ,,.
to measure
included in the @lendar
3 months);
6m.penods;
employment
that measure
sector at all
notion of low-wage
income, that arises from employment
quarters
(sigti&mg
variables
in the low-wage
a very gened
More refined notions of low-wage
analysis
anafysis:
pardcipates
only two time dummies
variables
in equation
and. withti
groups.
To
we simplify the regression
(3.1).
O“ne time dummy
is equal
to one if the period of observation begins in the y-s
1978 “ti ~983 and is zero
.
. .. .
,,,...,,.: ..... .. . ,, .:., . -..... . ,“. .’.
‘:’
:..
. ..’
..”’ ”.—.
.-’
othemise.
associated
Similarly,
the second time dummy
is equal to one if the time interval
with y, begins in the years 1984 to 1987 and is zero orfrerwi.se.
This fist
period is termed the late 70s and early 80s and we refer to the second period as the mid80s.
3.2
Participation
in Low-Wage
Employment
3-3
Table 3.1 -LQ and Table 3. I-M present tie pardcipation
markets
respectively
for the two definitions
each of the four education
of equation
i receives any mings
value of zero otherwise.
a dummy variable
from a low-paying
The upper en~
hbor
“LQ” and “M” of low-wage employment
The tibles report weighted
categories.
(3.1), with yi, representing
.,
irrdividud
rates in low-wage
for
least squares estimates
that ties
the value of one if
job in period t and it rakes on a
in each cell is tie fitted vakse for an individud
>
in the age range specified in the h~ding
represent
making
of the column,
tie fitted values from separate regressions
up our ‘sample--in tie order White, Blacks and Hispanics.
groups
The left group of
columns
reports results for the lak 70s and early 80s, and the right group presents
averages
for tie mid-80s.
answer
the questions
markets.
, ,,,.,,.
lower entries
for tie three race-ethnic
The results in Tables 3.1 -LQ and 3.1-M provide
,,,
and the tiee
posed in Section
1 concerning
tie information
participation
The main findings implied by the WO definitions
.,very sW,.
,,,. . . .. :;,.
permanence
of low-wage
categories
demographic
employment
by comparing
of low-wage
time horizons and age groupings.
of participation
to uncover
rates for different
how the composition
the predicted
The discussion
to
in Iow-wage labor
employment
The fo~owing &lscussion ~rst qes to gain Sorn= understanding
.....<.. ,... ..... ... .. .. . . .....:..~;..=:.. .... ... .,..:.
. .._..,, . .,. . ...
:... .=.. ... l. .
across different
comparison
necessary
of the
participation
rates
hen turns to a
ages, racial groups and edumtion
of participanfi
varies across these
characteristics.
Within education
categories
and calendar periods
3-4
a comparison
of tie predicted
are
.. .. .. .. ., ...,.
rates in the lower left hand OCUwith tie rates in the upper right hand cell suggests
participation
in the low-wage
sector is a temporary
event for most young men.
that
For
‘example,
in Table 3.1 -M during the lam 70s and early 80s, 75% of teenage high-school
graduates
hold a low-wage. job when considering
this group are employed
in the low-wage
we use a 3-month time horizon.
tables,
there is a signifimnt
term participation
participation
individuals
compared
While tiis example
sector is reinforced
mtegory.
yet only 11% of
reach tieir late 20s they are about half as Ikely
drop in the
This picture of short-
by examining
rates across both different ages and time horizons.
the patterns
k particular,
in
as
to hold a low-paying
job as
to when drey were in their late teens; and while one would expect widening
time frame more tian doubles
,,,. ,.,,.
ad
is tie most extreme
decline in every education
in the low-wage
horizon
sector by the time they are in their late 20s if
time horizon to lead to greater particlpatioi,
categories.
a 2-ye=
moving from a 3-morith horizon
the participation
the
to a 2-yeaI
fate for most of”the age-education
.
The vtiahon
in participation rates across different
...... .. ... . .. ,. . ,,, ,. :,..... . . .. . .. . . .
. ,,.
generaHy consistent
with the patterns displayed
TabIes 2.1; however,
and educatiomd
increases
ii education
low-wage
job
by the summary
Tables 3.1 point out some interesting
attainment.
statistics
interactions
As noted in Section 2, rates of participation
and yet a surprisingly
when &ey are in dleir early
Tables 3.1 reveals widening
demographic characteristics
. ... . . .,,
educational
large fraction
20s.
3-5
as individuals
age.
in
among age, race
decline with
of college graduates
A closer examiriation
differences
presented
are
,..
hold a
of the results in
I.n Particular,
for
men in their ~te teens the participation
the rates for high-school
tie participation
graduates
rates for high-school
and the some college group;
rate is higher for the dropouts
compared
Focusing .on the lower three entries in each cell suggess
among tie demographic
differences between
educational
category.
lower education
characteristics.
rhe pardcipation
rates of Whites,
For example,
fraction of Black college graduates
are very stiar
however,
to
for older men
to the other two categories.
tie~
is a subs~n~~
intemction
At the younger ages there are only slight
These differences
groups.
dropouts
are magnified
Blacks and Hispanics
k every
at the older ages especitiy
in Table 3.1 -LQ compared
.
for the
to Whites a siti
hold low-wage jobs in their late 20s, while at this
same age the participatiori” rate for Blacks in the other educational
categories
is roughly
double the rate for Whites.
In summary,
the signifitint
findings
revealed
in this analysis
we:
.
Involvement
in low-wage jobs tends to be a short-term
phenomena
and
is more concentrated
among young men. Cgnsequegtly,
measurements
designed to gauge the extent of low-wage employment
depend critically
on the time frame used to register participation
as well as on the age of
individual.
.,,
., ..,, . ... . .,,.. ..
.. . .
.
For high-school dropouts and high-school graduates,
participation
rates
in low-wage employment
are Iarge as 5990 when registering
participation
as holding any low-wage job in a 2-year period in the late
teens, and it is as small as 11% when participation
refers to holding a
low-paying job during an arbitrary
quarter when individuals
are in
their late 20s.
.
Regardless of the age and education of young men,
paying jobs is twice as fikely to occur over a 2-year
of the particular
quarters
making up this horizon.
●
Irrwpective
of the time frame
used to catalog
3-6
participation
in Iowhorizon than in any
participation,
the
?.:.....
.
likelihood of holding low-wage jobs decfines with age @y as much as
50%), with the most significant decreases occurring at the earfier agea.
.
3.3
.
A individual
age, education becomes a stronger factor determining
the
fike~iood
of holding a low-wage job. ~ile
participation
rates k 10Wwage employment
steadily decfine as individuals
squire
educatiorrd
attainments
beyond high school, a surprisingly
large fraction of college
graduates
hold low-wage jobs at the start of their wor~ng weers–
around 30~o in the first 2 years after graduation.
●
Race+thnic
differences in Iow-wage participation
are relatively minor at
young ages, but they tend to widen at older ages especiaUy at lower
levels of education.
Ignoring college graduates,
Blacks tend to have the
highest participation
rate, with no systematic
ordering between mites
andHupanics.
Relative
hportance
Tables 3.2-LQ
from low-wage
definitions
of Earnings
from Low-lVage
and 3.2.-M present estimates
employment
of low-wage.
of tile relative
to several sources of household
The tables report weighted
values imptied by equation
Jobs
(3.1) with y,, constructed
education
.
group.
Calculating
three ways.
the period
earnings
Calculating
y as dre
over his total labor earn~gs
. . .. . ,.:,
,.’
earnings
income in the period produces
the second. group of”rows listed for each education
ratio of an individud’s
of the fitted
.:
in tie first group of rows fisted for each
y as the ratio of an individrsd’s
jobs over total family nontransfer
of earnings
income for the two
least squares estimates
ratio of an tidividud’s
earnings, from low-wage employment
,.,,.
:..
. !:.,. ,, :. - ,. .,:$::’ ,,
in the period yields the results reported
contribution
from low-wage
yields a set of findings presented
category.
employment
in the tid
3-7
from low-paying
the estimates
Finally,
reported
~
calculating y as the
over total family income in
group of rows.
The upper entry
in each ceu is the implied
associated
fitted value for a nationally
with tie corresponding
lower entries present estimates
employment,
with separate
age-ducation
sample of youtis
category and period measure.
of the permnmges
regressions
representative
of income received
mn. for the three race-ethnic
The three
from low-wage
groups;
the first entry
refers to. White; the second to Black; and the third to Hispanics.
The results in Tables 3 .2-LQ and 3 .2=M portray
that low-wage jobs are an impotit
Further,
Witi
it is also a significant
a given education
category-calendar
income source over long
period,
a comparison
of the
overview,
as we~ as
upper left hand cell with the lower right hand cell provides a general
a ~ge
of values, indicating
determining
me economic
expect this diagonal
the irnpofince
as the width of the time horizon
considered.
horizon,
.,,
For example,
.
in the low-wage
these same income measures
obtain a signifiatfy
both short and extended
znd tick
families.
As one would
shows a decrease
in the impomnce
of”low-wage
window increases
and as broader
income measures
high-school
dropout’s
earnings
are
larger fraction
periods .of time,
in his late 20s derives only 38% of
from low-wage jobs,
across demographic
A similar comparison
of
groups suggests that Black high-school
of income from the low-wage
Specifically,
3-8
. .—:
,,,;
own labor income comes from
sector while a dropout
total family income. aver a 2-year horizon
dropouts
from low-wage jobs in
in .Table 3 .2-LQ for the mid-80s using a 3.-month time
.. ..... . . .
88% of a teenage
employment
of earnings
well-being, of individuals
comparison
suggesting
source of income for participant@ and ~eir famihes.
for some types of individuals
time horizons.
very similar pictures
over a 3-month
sector over
horizon over 85%
of a Black high-school
dropout’s
own labor income is derived
from low-paying jobs and
more than half (56-69 Y.) of his family income over a 2-year horizon
employment
in low-wage jobs.
A closer examination
rmderstiding
of young men.
measure
at different
fitie
income measures
of low-wage
In particula,
time horizons
work in both low-wage
the predicted
of particuh
of the importance
mobility
genedy
employment
exarninkg
provides
and tigh-wage
information
jobs.
wage employment
employment
25%-60 % of own Iabor..income
most low-wage
participants
with tie
is an important,
implication
significant
in bo~ low-wage
is much less important
is”derived
categories
and high-wage
from low-wage
tie estimation
employment
is
which indicates there is
jobs.
In con~ast,
employment)
sugges~g
in high-wage jobs over this period.
10W-
employment
There is an exception
group obti}n a.significant
Taken
to this
because a
fraction of
over longer time horizons.
results for the individual
results from the two family income measures
fiat
that low-wage emplOyme:t
.
men witi less than 16 years of education
number of men from this demographic
Comparing
using a 3-mondr time horizon
from low-wage
“event for most people.
own labor income from low-wage
labor income
on the extent to which individuals
results in Tables 3.1, these .rneasuresimply
.. . . . . ..:’
.,.
for Black and Hispanic
the economic
as the time frame widens to 2 years (e. g., only
are also working
but tempora~,
a more through
in determining
For example,
fraction of own laboi income resulting
“simultaneous”
provides
the total individud
above 70% for all ages, races and education
together
results from MS
provides
3-9
labor income measure with the
some insights
into the part low-
..,.:
paying jobs pby in determining
across dlfierent
less important
education
a family’s economic
categories
in determining
suggests tit
tie economic
circumstances.
low-wage
circumstances
This comparison
employment
is relatively
for tie families of men with
higher levels of edueation.
For example,
typical high-school
in fie mid-80s who ‘is in his early 20s receives 85% of his
dropout
own labor income versus 73% of toti
comparable
ftiy
college graduate
income
low-wage
and longer time hotions
entries
jobs.
less..irnpo-t
in detefifing
are c~rrsidered.
employment
for Whites,
Black men with lower edumtional
earnings
Similar comp~i~ons
Conducting
in each cell point out some significant
low-wage
from low-paying
hotion
family income from low-wage employment
receives 71% of own kbor
from low-paying
jobs are relatively
in Table 3 .2.LQ using a 3-month
differences
whites-
a
[email protected] Kqt earnings
from
ss men age
the same exercise using the lower
levels (i.e., categories
.!o
w~e
income and only 54% of toti
a family’s’ ticome
Blacks, and Hispafi=-
jobs relative
a
Thgss
in the relative impo~ce
~ Paficular,
of
the families of
11-; 12 and 13-15) rely more on
racial
diffe!enc.es
are,.not
as
prominent
for college graduates.
,.
,.,. ,..
: ... . ............ . . .. ...... . ... ... . . . . . ..
.
. :. ..-.-,..... . ... -.,
,,
The principal conclusions drawn from “his an~ysis on the relative impo~rrce
of
earnings
from low-wage jobs are as foUows:
●
The fraction of income made up by earnings from low-wage jobs falls
as one considers longer horizons to accumu!.ate income and broader
sources of income.
Earnings from Iow-wage jobs is a relatively h]gh
fraction of individuals’
total labor income earned during the quarter in
which these jobs are held; it is a much smaIler fraction of total family
income received in the 2-year periods inchsd]ng these quarters.
●
For high-school
dropouts
and high-schooI
3-10
graduates,
earnings
from low-
.
wage jobs average as much 86% of total labor income during those
quarters when such jobs are held; this average drops to as fittle as 30%
when maculating the contribution
of these earnings to total family
income received over 2-year periods during which low-paying
employment
takes place. For college graduates,
the analogous
percentages
are 71% and 15%.
●
Considering
individual’s
labor income alone, the fraction of earnings
coming from low-wage jobs for Iow+ducated
partieipanfs
ranges
between 69.% and 88% when. considering
the contribution
of these
earnings to quarterly
labor income; and it ranges beween
3390 and
61% when considering
the contribution
to totai” labor income received
over 2-year periods when low-wage employment
occurs.
●
For poorly edumted men who hold Iow-wage jobs, earnings from these
jobs on average account for 65%-80% of total fmily
income received in
a quarter,
and 40%-5070 of family income received over 2 years periods
when low-paying jobs are held.
●
me contribution
of earnings from low-wage jobs to any measure of
income over any period declines with age and with higher IeveIs of
education.
me rate of dectine associated with aging is much steeper
for the upper education groups.
●
At the Iower education Ievels, the fraction of any measure of income
made up by earnings from low-wage jobs is typica!ly highest for B1acks,
with Hispanics second, and with mites
having the lowest fractions.
Race-ethnic
differentials
are insignificant
for college graduates.
✎✛✌
✎✌✌✌✌✌
... .,..
. ..
... .. . .. .
.,,
3-11
—
4.
Understiding
extensive
information
An Empirid
Frmework
the process underlyirig
economic
on the patterns
work experiences
of heir
mobility
for a population
from the time of initial
entry into tie labor force until a distant point in tieir
working
longimdind
the early potion
data on a sample of individuals
covering
careers.
say at least he first ten years after leaving school for each person,
fitie need to introduce a statistical
paper.
tie toti
amount of time that persons
akemative.
horizons,
persons
occupy in different
Unfo~nately,
cycle or. dendar
,.,
substantially
difficulties
there would be
later in this
statuses over
by individuals
to employment
Invfiably,
for shorter periods--~~eed,
. .
,,-.
the long horizon
of inter%t ii–his
. .
study; and
which implies hat no
is available on their work ~listories. from”-the. time &at they depafl
school unti when they become sample members.
conclusions
labor-market
enter the sample after they have left school,
detailed information
hen
over the same period on either tie life
time); ~ey are .Wpically obsemed
,,, ,’...
. ..
shorter in many instances--than
some individuals
of their Metimes,
such dati are neve~ avaiIable.
included in a data source are not observed
.
. .
a variety of statistics summarizing
along with the various routes fo~owed
paying different wage levels.
If we had detailed
model to infer tie findings developed
Given such data, one could readily calculate
requires
These complicatioris
that can be drawn from simple summary
in any empifid
analysis
of economic
4- I
measures,
behavior
severely
hmit the
and they create signifiat
in a dynamic con~xt.
-
As a conceptual
introduces
expanded
framework
a generali=d
for dealing witi
variant of.a transition
beyond that found in applications
flexible econometric
labor markets
framework
youtis’
experiences
Iabor-market
The perspective
Our aim is to describe
stages of ti”dividuals’
chains.6
the sequences
The resulting
adopted
substantially
This model offers a
and movements
among
to and fro[n educational
time--as
the patterns of experiences
amount of time that persons
analysis
in tie model is to characterize
using age--not calendar
lifetimes after they leave formal
focuses on the cumulative
the relevant
tilne
that occlir over alternative
schoofing.
interest
not only
spend in various activities,
it also
of these experiences.
empirical
model characterizes
wage labor markets and of earnings
.,
of Markov
model CPM),
paying. different_ wage rates, as well as movements
activities.
considers
probability
the subsequent
for studying economic. mobility
and nonemployment
measure.
these issues,
six aspects
of participation
in low-
mobility in general:
●
The probability that a youth will enter low-paying
portions of his Iifetiine.
employment
●
The likehhood of entering or returning
alternative labor-market statuses.
●
The lengths of spells spent in low-wage employment,
in high-wage jobs, in nonemploymerit,
in educational
combinations of “these statuses.
to low-wage
employment
over various
from
as well as in employment
activities, and in
. The likelihood that a low-paying employment spell will end with a participant
moving to high-wage employment, educational activities or nonemp[oyment.
6 For further discussion of elementary variants of Markov chain models, see the textbooks
by Bartholomew (1982) and Howard (1971).
4-2
.-
. Measures of the toti time spent in low-wage employment--or
in any otier
-over an extended period of time, including the number of spells and the
cumulative duration during a specified horizon.
. The way in which experiences vary among
characteristics
and labor market histories.
We formulate
on a wee~y
the TPM to characterize
basis across the five econolnic
status t (employment
job);
in a low-wage
status b (simultaneous
(involvement
in some type of training
and not participating
in educational
of the hourly wage distribution)
low wages in assigning
m~
Out
probabilities
alternative
describe
uninterrupted
that individuals
routes).
characteristics
activity);
We consider
in a high-wage
job);. StatUS e
and status n (not working
both the LQ (lowest quintile
incovomtes
three-basic
and entrance
probabilities.
that individuals
wage) definitions
of
The entrance
the analysis we account
statuses
for porential
and the possible effects of history
4-3
The tiitial
career in
The duration
of w~eks that individuals
probabilities
a spell enter alternative
buildingbloch:
stirt their working
at the time that they leave school.
the length of spells or the number
who terminate
mobility
in Section 2, which include:
status h (employment
or educational
framework
state.
individuals’
for these. two definitions.
the likelihood
in a particular
Throughout.
defined
different
to state ? ii a given week, which means that we
analyses
labor market activities
distributions
governing
in both a low- and a high-wage
duration distributions,
summarize
possessing
and the M ($1 above the minimum
The subsequent. empirical
initial probabilities,
statuses
pursuits).
an individual
.Separate empirical
tie process
labor market);
elnployment
individuals
state-
variables
summarize
spend
the Iikefihood
fi.e., leave via alternative
differences
in individual
on the various statistical
relationships.
model,
witi
empirical
4.1
Our statistical
model provides
many features incorporated
applimtions
Probabilities
elsewhere
Wtermining
is tie sWcification
upon finishing
Initial
specification
specify tiese
as
terrninglogy
labor market
probability,
over their
enter immediately
tiis. conlPonent
mobility.
dete~ines
The statistical
for it reflects the Iiketihood
We
.
Pr(s+i)
for the probability
=Pr(s+i[
X)
i = t, h, b, e, n
that a person exits from school (smte s) to state i as the first labor-
at tie end of school.
that an individual
analysis,
Formally,
the quantity
Pr (s + i) represents
moves from state s to state i conditional
demographic
way in which various
subsequent
work_ experiences
enters each of the states f, h, b, e, o~ n at the end of school.
(4.1)
By inco@orating
Status
status that persons
In econometric
of dre process describing
probabilities
market statis
which are not found in
individuds’
of the labor-market
needed here is a type of entrance
that an individual
probabili~
Labor-Market
to characterize
formal schog!iflg.
the initial conditions
in our specifications
to a competing-risks
in &e literature.
The first “elementneeded
lifetimes
a flexible alternative
segments
X includes
characteristic
on the covanates
in X, one can allow for differences
of the population
variables
the
reflecting
4-4
start their working
schooling
careers.
attainment,
in the
In the
”age, and race.
X.
4.2
~ration
Distributions
The next element
and H=rd
quired
to characterize
of the length of time fiat individuals
distribution
characterizes
weeks of continuous
residence
labor-market
spend in the various
the I&eIihood
experiences
is a summary
statuses upon entry.
fiat an individual
in a particular
tiis status and some specification
According
~t=
experiences
A duration
a given number
labor market activity given admission
of
into
of history prior to the start of the spell.
to our formulation,
an individusd may occupy any one of the five
smtoses
(e, h, b, e, or n) at any given time.
duration
distribution
htting
“i” designate
an arbi!raw
S@US, a
takes the form
(4.2)
J(7)
== si(T-l)
H,(7)
,
witi
(4.3)
Hi(/)
=
l-Pi(r,
z3
and
(4.4)
S,(T-1)
In these expressions,
dre variable
market status i, and Z accounts
of-these spells.
remains
weeks,
.
The probability
7 corresponds
Pi(t,
z)
~
to the duration
for factors other than duration
Pifi,Z) designates
in the current Iabor,market
with the covariates
=
,-1
~
,-1
the likelihood
of a spell spent in laborthat influence
tie lengths
that an individ.llal
stahts after already having been in the state for t
Z summarizing
the demographic
4-5:
characteristics
and previous
work history at the stafi of current episde;
The finction ~(?) specifies
for individuals
spell.
tie Iikelihod
falling into a mtegory
the quantity
an episode
function,
particular
since leaving
school;
to entering
of tie low-wage
4.3
participation
of the labor-market
his current position.
individual’s
The tiird
who, having been in
Finally,
that an individud
in our arialysis:
will
the components
H sun]rnarizing
in H, including:
experienced
of whether
of training
in the year prior to the
the individual
completed
participated
are interacted
in any
since leaving school;
status from which tie individual
Most covariates
an
total employnlent
in each of the two years preceding
spell in the current labor-market
Entrance
of the
prior to the start of the current spell.
measures
spell; an indicator
since leaving school; the amount
spell; and indimtors
the probability
and the variables
the amount of employment
of tie current
indicators
Z at” the beginning
of the population
cornponenk”
work history experienced
a wide variety of summary
training
depicts
characteristics;
We consider
initiation
tie fraction
Z include two distinct
for.. demographic
individual’s
t and Z.
in SQNS i that lasts at least r-l weeks.
The covanatei
X accoundng
by attributes
status for t-l weeks, will leave on the t-th week of their spell.
S,(T-I), the survivor
experience
on the variables
dlat a spell in status i will last exactly 7 weeks
characterized
The hazard rate Hi(t) determines
tie labor-market
Pifi,Z) conditions
the current
exited just prior
with t, the length of the
status.
ProhabiUties
element
needed to characterize
4-6
labor-market
mobility
summarizes
the
Welihood
hat
of a spell.
enter the alternative
This component
demographic
available
individuals
of the model determines
characteristics
to persons
labor-market
influence the prospeck
statuses
how different
of following
upon tie termination
work histories and
the various paths
when exits occur from each status under a variety of circumstances.
Upon the conclusion
of a spell in state k, define the probability
that an individual
enters state i as
(4.5)
Forma~y,
Pr(k+i
Pr(k+i)
=
the quantity
Pr V -i)
state k to labor-market
\T,
represents
Z which describe circulistrmces
Incorporating
work history variables
for entering a new labor-market
demographic
groups.
Estimated
information
anY horizon,
characteristics
moves from
in Z-allows one to
wo”rk experiences
stares and how these prospects
influences
differ across
Experiences
variants of the preceding
statistical
needed to describe events associated
specifications
provides
all the
with time spent in low-paying ,jobs over
along. with the pattern of transitions “to and from slieh employment.
this knowledge,
experience,
that an individual
nature of each individual]’s lifetime
Employment
h, b,e, n .
at “the start of the spell that just ended.
and demographic
the prospects
Characterizing
i=t,
on ending a spell of 7 weeks in state k and on
the covanates
4.4
i#k,
the probability
activity i conditional
model how tie particulm
Z),
one can infer a wide array of”dimensions
such as: the probability
associated
that a youth will enter low-paying
4-7
Using
with low-wage
employment
over
various portions
number
of his hfetime;
the routes of en~
of entries into low-wage
jobs; and measures
of the tod
The above elements
charactertiing
calculate
the appropriate
probabfiitiei;
member
time spent in low-wage
patterns.
the Mefihood
product
of a demographic
employment.
blocks for several approaches
W bile not the most convenient
of any particular
of” initial probabilities,
such a procedure
jobs; tie
the lengths of spe~s spent in low-wage
serve as building
employment
analytically
employment;
into and exits from low-wage
employment
duration
for
approach,
sequence
distributions
by forming
and entrance
literally permits one to predict the likelihood
group will experience
one can
any specific pattern
that a
of labor market
experiences.
To iUustrate this procedure,
and the survivor
define f~i(~~ and S~i(~ti-l ) as the duration
furiction associated
with the ith spell experienced
state k, where k = t, h, b, e, and n, depending
wage job, .a high-wage
@aining positions;
corresponding
sirnultanegus
or nonemployinerit.
expression
entrance
probabilities,
state m.
Relation
quantities
job,
on whether
employment
by an individual
the person
in low-wage
an:
occupies
high-wage
Each fti and Sti has a form analogous
listed in (4.2) and (4.’4). Similarly,
distribution
in
a lowjobs,
to the
let Pr(ki + mj) denote the
linking transition” from the ith spell in state k to the jth spell in
(4 .S characterizes
the form for these probabilities.
f7-S, and Pr in a way that describes
an individual’s
Multiplying
experience
provides
the
the
.
expression
for the probability
associated
with, tl~e particular
to this experience.
4-8
speu sequence
corresponding
For example,
person
the likelihood
starts a spell in low-wage
that, over a period of T“ weeks after leaving school,
employment
at that time, and spends the remaining
If, instead,
this individual
to the end of the period,
nonemployment
probability
(4.7)
Pr(s-e,)
where
completes
is given by
S.1(7.1)
the spell in the nonemployment
after ~a, weeks prior
for 7~1weeks, and again enters
~ti weeks which runs to the end of tie period,
then the
is
ftl(r,,) Pr(el+n,) L, (+,,,) Pr(nl+hl)
&l(Tkl) PrfiI+nd
S,,~(T”J,
~fl== T.r( ,-~., -7~,.
The use of analytical. methods
dynamic
patterns of Iabor,market
infernng
the likelihood
straightfo~ard,
amount
f~ed
moves to nonempIoyment
7., =T-Tfi weeks in nonemployment
enters high wage employment
for the remaining
implied
~t, weeks,
ft,(rt,)
Pr(el-nl)
Pr(s+P,)
(4.6)
bsting
a
for calculating
experiences
of any particular
it is exceedingly
of time that an individual
statistics
ii’ computatiorially
sequence
complex
summary
of experiences
to calculate
can be expected
burdensome.
While
is relatively
such statistics
to participate
describing
as the average
in a specific””activity
over
horizons.
Simulation
methods,
offer a far simpler approach
labor-market
experiences
horizons
activities.
of the sort now commonly
for computing
The subsequent
of the four ed~lcatiorr groups
used in statistics
suInmary
s@tistics designed
analysis utilizes
considered
4-9
to characterize
such methods to summarize
in the previous
covered in the first 10 years of their working
and econometrics,
discussion
during
cafeers after leaving” ichooI.
5.
kplementirrg
Empirical
the empirical
estimation’ of three statistiml
probabilities
Pi(t,Z),
en&ance probabtities
statuses
quantities,
which determine
outlined
considered
in Section
Pr @ + i), which indicate tie likelihood
specifications.”
describes
for these quantities,
The discussion
the specifications
tie implications
Proposing
and results in greater detail,
these specifications.
the various
~lstributions
alternative
Pr (s + i),- which designate
activities
specifications
ody
after leaving
adopted
implemented
briefly;
Appendix
in our
to estimate
B
and Section. 6 filly characterizes
and H=ard
for~ and Hi requires
the probability
z given characteristics
the
findings.
of Duration
specifi=tions
Pj@,Z), representing
results
activities;
of entering
along with the procedures
o“titlines estimation
of the empirical
5.1 Pararneterization
the empirical
the
in the order: the
the length of spells in labor-market
This section introduces
analysis
4 requires
nlost conveniently
of srarti.ng a career in .rhe vario.tis. labor-market
schooling.
empirical
duration
framework
and Estimation
at the end of spells: and the initial- stahls probabilities
the Ikelihood
fonnd
Specifications
of remaining
Rates
an appropriate”
types of work activities,
the nature of duration
which primarily
5-1
form for
in stite i for one more week beyond
Z. Three aspects of his function+
The first involves
filnctional
determines
..
form are critical to
dependence
applicable
for
how P varies with t.
The second concerns
are included
the effect on P of an individual’s
in the covariates
of duration
dependence
possibility,
we require
form for duration
different
a fom~ulation
empirid
specifications
dependence
for P that admi~
dependence,
and covariates
flexibility
dependence.
mode for presenting
for such a
boti in tie functional
vanes
for
information
about the
Graphs of hazard rates in our data reveal that
of the probabilities
specifititions
features
-_
typically
Pi(f,Z) must admit nonmonotonic
duration
to vary according. to the attributes
do not accommodate
nor do they allow for interaction
between
nonmonotonic
Z.
duration
tie form of “duration dependence
Z.
To allow for such flexibility,
probabilities
To account
which
Z.
and rdlow the form of his dependence
empirical
that central
and in the way in which this dependence
hazard rates is a popular
of duration
Stindard
change as the values of Z change.
dependence,
chamcter
work experiences,
Z. ..The third relates to dre possibility
values of tie covanates
Plotting
previous
we specify the following
logistic modelfor
the
Pifi,Z):
(5.1)
Pi (Z, t)
=
1
~~ ~z,P,+8’(f.q>a,)
‘
-—.
where .ZI and Zz are vectors of variables
vector, and the finction
made up of the covariates
gi ~, Zz, a) determines.
time spent in labor market .smtus i.
Z, Bi is a parameter
the dumtion’ properties
ass-o”c-iatedwith the
The covariates
variables,
market
Z include demographic
H, where dl of these variables
episode.
variables
The characteristics
evaluated
fi.e.,
representing
the groups
logarithm
of tie total number
m~sures
the person’s
education
of weeks
tie 52-week
signifying
wl]efier
he, individual
preceding
the current spell; tie total number
product
of this proportion
individual
the proportion
left school (i.e.,
the log of potential
tie economic
indicators
In specification
signifying
whether the
the two years
since
of the total time since” leaving school;
of the total number
between
the fraction
prior to entering
for school s and for the five labor-market
of ““totil employment
variables
the current
statuses
and
signifying’
status--
P,. h, b, e, and n.
gi(t, 2?, a) not only Caphlres the properties
5-3
.,.
the
of weeks since the
and a set of 6 indicator
by the individual
(5.1), the function
during
variable
of weeks spent in training”” programs
labor market experience);
status occupied
variable
job for _at least 1 week during
and the logarithm
of
in a low-wage job for at least 1 week
spent employed
an interaction
with indicators
the fraction of time spent employed
spell; an indicator
was employed
school;
and the natural
of the current spe~; an indicator
individual
leaving
for educational
analysis.
was employed
in a low-wage
indicator
left formal schooling--which
is fully interacted
H include:
tie current
variables
since the individual
period prior to the initiation
the year preceding
including
indicator
levels, as is done in the subsequent
variables
at the start of the current labor-
11-, 12, 13-15, and 16+);
age when .tiis variable
The work-history
X, and work-history
making up X include: an intercept,
for racial origin [email protected] and Hispanic);
attainment
during
characteristics,
of
duration
dependence,
according
to au the attributes
for modelling
of duration.
duration
hplicit
series of intervals
nondifferentiability
effects,
variable
which fit polynomial
smoothness
requires a ticrifice
of detecting
to vary
approach
h smoothness.
functions
complicated
=
to a-
of fit.
.—
but
Limiting
yieIds a smoother
the
cu~e
but
forms of duration dependence.
,*”[~~(f)
‘i,j-,
functions
and goodness
functions,
we specify gifi, Zz, a) as the general
@~(f) denotes
possessing
are an attractive
the number of polynomial
or the order of the polynomial
gi(t> z2, ai)
The quantity
spline models,
is a tradeoff between
by increasing
In our approach,
dependence
since they fit the data with a flexible and smooth function
in conventional
the capabilities
(5,2)
in Zz. Spline models
at the boundaries
of intervals
diminishes
of Zz in gi also allows duration
included
over duration,
Flt can be improved
number
but the presence
‘~ + r a;j~]
‘r)][a/jo
the cumulative. distribution
function:
functi.oi
of a normal random
mean pi and variance ai~; and the ailj’s in (5.2) represent
parameter
vectors.
The presence
of Zz in g, allows duration
attributes.. includ–ed in ZZ.., TO ~qqribe
that 2; only consiss
of an intercept;
the cd~s in (5.2)” permits
polynomial
suppose
dependence
the c!!ar.a-cterisqgs. of
so av~ Z2 + z a;j,
gil.
=rrio
us to “inctirporate spline-feamres
atio + a~l t represents
to vary according
g over only a specified
?uPPose
for.
to all the
fig
moment
+ mi, t . The presence
of
in g so that the linear
range of r. In particular,
we wish to set gi = ai~o + ai,, f for values off between O and r“ and to set g =
5-4
aizo + aizl t for values of f between
t. and some upper bound f. To create a
specifimtion
of g hat satisfies.. me property,
determining
the cdfs as pie= 0, p~l = to, pi~ = f;. and pick very small values for tie
three standard
tie quantity
assign J = 2 ~ (5.2); fix tie Aree means
Oio, ail, and ui2. These choices for the p’s and the a’s imply that
deviations
over the range (0, t“)and =0 elsewhere,
@i,fi)-4io(t)
= I
and tile quantity
-,
@i,(t)-@ii(r) = I over the range (f., i) and O elsewhere.
desired property.
Further,
gr(t, Z2, a) is differentiable
and the u; set in advance of estimation,
and in known
functions
oft
~ifi, Zz, a) is sticdy
Me values of the p’s.
each sphne cuts in and out by adjusting
5.2
for a. more gradud
Estimation
of Duration
We estimate parameters
specifications
~SY
analysis
supplemental
likelihood
observations
irrfonnation
the
tie values of the pi
linear in tie parameters
and smoother
transition
Distributions
and Hazard
of tie transition
to estimafe
samples
u
maximum
included
~(7).
methods
among our work-histo~
in
of tile sort
The incorporation
of the
maximum
Our sample consists
our period of observation.
5-5
to the next.
P@,Z) appearing
likelihood
of tiese probabilities.
on all spells initiated during
values
Rates
probabilities
tie distribution
how quicwy
from one polynomial
into our data set requires the use of weighted
@chniques in the estimation
on the variables
One can also control
the values of the u’s, wifi higher
(5. 1) and (5.2) by implementing
found in duration
in f. Witi
g, possesses
and Xz. One can control where each spline or polynomial
begins and ends by adjusting
providtig
Accordingly,
of
We have complete
covanates
H, so we
encounter
no problems
We estimate
when spe~s are interrupted.
categories,
and our two low-wage
In estimation,
of the empirical
by hcluding
specification.
transition
d
exploratory
analyzed.
data analysis
dependence
The empirical
that determine
parameteri=tions
estimates
the formulation
the five duration
for choosing
(Appendix
and the characteristics
for each of the
specifications,
acrms
B presents
category.)
we adopted
the alternative
hypothesis
tests indi~te
Regarding
of the spline functions,
that sigpifi~ant
labor-
the details of the final
the namre of
tie values of the
and a‘s vary across spell types, but are held constant across education
bnventiorrd
from tie
distriblltions.
of the tra!]sition probabilitia’
categories.
of tie
to vary with all covariate.s
the a coefficients
for each spell type and education
dependence
We selected the values
of Z listed above in Z,, with Z~eliminated
analysis
witi
for each type of spell; the
We allow tie form of the duration
market staruses and education
duration
stabs
by setting J = 5 in (5.2),
determine
probabilities
specifications
of gi(t,Z2,aJ
hazard rates principally
the components
for right censoring
definitions
Fo~owing. some general guidelines
different
accounts
modeIs for each of our four eduationd
to the particular
for the pk and a’s after extensive
function gio.
distinct
we pick a specification
the p’s and u’s freed according
properties
Our procedure
with left censoring.
p’s
categories.7
interaction. effec~ exist be~een
7 Extensive tesfi~g ind~ti”tes that th”e null hypothesis. of ayl = O”for combinations o-r
inditiiduti valum of i an-d j cannot be rejected at converitfonal significance” levels; thus, the
estimated transition probabilitim” used in the subsquent analysis impose this restriction on the
parameters in quation (5.2). Furthermore, after testing, we restrict the ai~o = O in estimation,
except for the coefficient on the intercept term in ~; thus, the transition probability is consmnt
and does not depend on individud characteristics beyond the last spline point.
.
5-6=
duration
and demographic
characteristics
and between
duration
some range of weeks for all of the spe~ types and education
tie components
5.3
of both X and Has
Par~eterizations
of Entrance
The probabilities.
.
individual
k and on the covariates
demographic
activity i conditional
how various circumstances
determine
thus, we included
in ZZ.8
on ending
related
the likelihood
~ and Z influence
to previous
Pr. k + i) are also likely to display
To offer a flexible specification
analogous
the value of
work experiences
and
exit from labor-market
history dependence.
for entrarice probabilities,
using a multinominal logit specification
hat an
a spell of 7 weeks in status
affect the route by which individuals
status k. The probabilities
quantities
included
Z. Tile way in which the variables
characteristics
=tegories;
over
Probabilities
Pr k + i), defined by (4.5),
enters labor-market
Pr @ ~ i) specifies
part of the covariates
and history variables
We’pararnetenze
these
to the formulation_proposed
in
(5. 1), which takes the form:
(5.3)
Pr(k
+i)
=
Z,D,,+g,,(r.z,.ni,)
c.
zL6ti.K,(7,&,m*)‘
j.t,E.e.*
i~k,
i=t,
h, b,e,
n ,
e
where ZI and Zj are vectors .of variables
Imade up of the covariates
2, ~ii is a parameter
8 To simplify the model, we restrict the coefficients o“n the vtiables in Zz, other than the
intercepts, in two ways: first, if none of the coefficients are individually significantly different
from zero at conventional significance levels, and a joint hypothesis test indicates that we a
accept the hypothesis that dl of the parameters in a spline are equal to zero, then we impose
these zero restrictions; and, second, we constrain the coefficients to b.e equal across spline points
whenever a joint hypthesis test indicates this is a valid restriction.
5-7
vectoq
and the finction
gti(T, Zz, a~) determines
how the l~eiihood
of various
entinces
changes with the length of tie SPI1 hat has just terminated.
The covariates
variables
Z include tie same demographic
listed in Section 5.1, in addition
any training
characteristics
to an ““indicator variable in H signifying
to interact with the duration
dependence.
We parametrize
h (5.2); the quantities
distribution
function,
of the previous
SIi(.) similarly
as before,
in modeI (5.3) allows
spell, thus controlling
to the overlap-spline
@{j(T)in this formulation
represent
and 7 is the spell Iengti
for all i and j, the model can be interpreted
over which the dummy
effects as duration
cotsld detect relatively
particular
variables
satus.
overlap-spline
apply.
changes
high tendency
This phenomenon
specification
allows
gradually_ witi rime, according
of the previous
In this specification,
for the spell just ended.
for individuals
Estimation
of Entrance
Probabilities
5.8
gi(.) specified
spell.
If qj = O
variables
in
de~rmine
he range
the ~-. v~iables
can have
For example,
in the a parameters.
for the effect of the dummy
to the properties
for history
the model
who have short SW1lS to enter a
would be reflected
.
nomal
as interacting. a-setie$ of dummy
@iJ(r).
5.4
function
the cumulative
duration. w.i.th the covariates ‘“inZ-, where. the values of the ~~jte~s
different
whether
took place before the current spell.
The form that we specify for the function gii(7, Zz, a~
covariates
and work history
variables
of the cumulative
Our
to change
distribution
functions,
The parameters
likelihood
techniques
As with the duration
nonrepresentative
of the entrance
using the multinomiai
specifications,
categories
sample consists
Iogit specifimtion
Our estimation
specifications
on completed
of&e
Z, incorporate
H in Z, and specifies
weeks since the individual
employment,
number
left school,
the six indicator
term and long-term
nonlinear
m=sures
episde
measured
variables
for race-ethnic
of the logarithm
variables
for the preceding
of the fraction
in equation ‘C .3) that
gii(T, z~, at) as a finction
the two dummy
of weeks spent in training programs
since he left school--all
probability
indicator
functions
variables
for the
spells.
entrance
length of the current spell with L- includ~g
1
(5.3).
and each of the five labor market statuses.
tie history variabl~.
The covariates
in eqmtion
for each of the
Distinct
procedures
appearing
maximum
are estimated
of all observations
weighted
by weighted
to account
We adopt a parame~rimtion
includes
are estimated
are required
sample design,
four edumtional
probabilities
of the
groups.
of the total number
for previous
economic
of
.1
low-wage
statis”; both the short-
of time spent in employment,
the total
and the indicator ..variable for any training
at the time the individual
started the labor market
that he is just leaving.
The formulation
of the function
the p’s and u’s acco~ding to. the specific
dummies
in ~.in
the likelihood
gliO used in the estimation
smms analyzed.
Inclusion
g~iO means that the length of jllst completed
of entering
Blacks, and Hispanics.
tie alternative
The empirical
labor-market
analysis
5-9
s~~=s
estimates
sets J =.2 and picks
of ra~e-ethnic.
SPIIS in s~hls” k influence
differently
for Whites,
the Q and a mefficients
that
“1
I
determine
5.5
the four entrance
Parameterizations
The empirical
natural candidate
likelihood
probabilities
for each of the five labor market statuses.
of titial-Status
formulation
for specifying
of entrances
in~oduced
above for tie entrance
the Pr (s - i), for fiese
into” the various
work career upon the completion
rules out the possibili~
Probabilities
labor-market
of school.
probabilities
statuses
The nature
of tie probabilities
demographic
characteristics
in equation
also measure the
of these initial-status
(5.4) depends
is a
at the start of a.n in@vidual’s
that they depend on any work-history
specification
probabilities
variables;
exclusively
.
probabilities
and, tius,
the
upon the
,
We parametrize
Empirical
formulations
these probabilities
using”.a multinominal
Iogit specification.
take the form:
Pr(s
(5.4)
X.
+ i)
i=t,
h, b,e, n
,=,,;:ex5.3-’.””
.
The covariates
5.6” Estimation
X include tie demographic
of Initial-Status
characteristics:
age and race.
Probabilities
.—.
We implement
specifications
of the empirical
maximum
likelihood
(5.4), using weighted
analysis,
techniques
procedures
to account
there is a single observation
5-1o.
to estimate. the parameters
for sample design.
of
In this step
for every young man” making up
our sample because
the observation
the selection
period.
k the specification
and several terms designed
influences
entrarices
by incorporating
elements
of the covariates,
we estimate
X in (5,4) includes:
to allow for a flexible formulation
components
labor-market
statuses.
of a vatiant of tie finction
of X, with the argument
different
mtiels
the mce-ethnic
for
We construct
these terms
gfi, z?, a) specified
(5.4),
groupings
The parameterimtion
intercept.
5-11
in (5.2) as
in montis)
of go used in the empirical
with the p’s and u’s in the functions
and with ~- containing
dummies;
for the way in which age
“t” defined as the age of the man (measured
schooling.
sets J = 2 in equation
across tie four educational.
to Iwve school during
~tegories.
into the alternative
when he exits from formal
analysis
requires the individud
To admit flexible specifications,
each of the four education
.
criterion
nothing
@;jdiffering
~nore man an