PDF 1

An Analysis of the Consequences
of Employer Li&ed
Health
Insurance. Coverage in the U.S.
Eric P. Slade*
Brown
December
U-niversity
.21, 1995
*This project w- fmded by the U.S. Department of Labor, Bureau of Labor Statistics under grant
number B-9-J-4–3509.
Opinions stated in th~s document do not necessarily represent official position
or policy of the U.S. Department of Labor. I gratefully acknowledge the tremendous benefit to this
project added by the comments of Tony Lucaster, Rachel Wledberg, Wbert Moffitt, and workshop
participants at Brown University and the Bureau of Labor Statistics.
EXECUTIVE
The
Consequences
of Employer
SUMMARY
Linked
Health
Insurance
Coverage
in the
for whom job
chages
U.S.
Eric Slad.e
h
the U.S. labor mmket it is commonplace
sdted
in timeases
and ch=ges
comody
jobs
in hedtb
insuace
coverage
that do not offer health insumce
job receiving
their t enme,
health
insmance
coverage
The usual mrmgement
poficies
to employers.
ksof=
as tidividuals
level of health insmance
and Mofieit
E&ed
years at the s-e
(1993)).
tiom.
recomend
set of proposals
medical
mdyzes
offers new evidence
with a lengthy
&=ges
ad
is stronger
regtiation
traditions.
Some work-
othm
workers l-ve
as a benefit
to ctige
tboughout
offer of
employers
in the U.S. today E&s
coverage.
reduce
‘job-lock’;
is tiefficient,
fided
of job
that employer
chmges
md
with pre-tisttig
medicd
health insm-ce
coverage.
from -eluding
a new employee’s
set of propos~
the s-e
cwent
(1994] and Cooper
they conclude
the frequency
for individuals
A second
oft-
health insmance
that this amagement
ti a new job (e.g. Mati]m
compties
yeas
employer.
of employer
the option of contenting
at his fomer
This paper
u
have su~ested
be reduced
wotid prohl~]t inswanm
offer a new employee
job
inaeased
cqnd]tilons from poficy
receivtig
be observed
coverage
does substtiti~y
effect of health immmce
that individuals
to work irr preferred jobs due to a fear that thei
coverage wodd
the negative
revd
job ordy to have theh employer’s
fight
through
a few economists
cmerage
co~rage
coverage plm
These authors term the tiefficiency
health imwmce
They
StiU others
pass up opporttities
wodd
for ones that do not, whle
insm-ce
for health insmance
Recently,
by individuals
for ones that do. Some workers work ~ny
health
coverage
have re-
Simple impect ion of the job chmges
health tisuace
coverage
withtiawn.
never receive health insm=ce
coverage.
coverage
thek employer’s
wtile others work may
hdlviduals
HP erienced
different types of job ad
jobs that offer health insurance
in the s-e
-d
or decreases ti health insmace
e~erience
ws l=ve
to observe
health insmmce
wotid
requke
poEcy
that
con&One
pre--sting
-ployers
whid
to
he had been
employer.
the mgments
regmding
the tiect
critique of etisting
health insmmce
=d
evidence
of health
studies.
coverage.
immmce
The second
The fid
ES-1
presented
p=t
in efisting
covwage
job-lo&
on job mobfity.
of the paper presents
section
reports
studies,
the resdts
and
It begtis
a new model
of m =piricd
of
analysis
based on data from the National
Existkg
studies
of the effect of health insmmce
view of the relationship
that m individual’s
is connterfactual.
chmacteristics
addess
between
coverage
experience
of m individual’s
change.
in etisttig
The empirical
coverage
in etisting
differences
health insm~ce
insmance
allowing
job-lock,
correlation
caused
insmmce
coverage
Thee
questiom
chmges
heath
md
as a benefit
that job mobility
insmmce
of tigfdy
Recognition
Etisting
papers
of
do not
They dso assue
employer
is rel=ed
the job-lock
health insmmce
affect a
propensity
these ttiee
both the frequency
They may prefer jobs
mobile,
or they may
Either way, such higfdy mobile in@viduds
at jobs
the mtin
individuals
which do not include
empirical
resdt
coverage,
into jobs whi&
section of the paper:
job mobi~ty
for job mobitity
health
in studies
of
may be a spurious
do not offer health
into jobs. which do offer health insmmce
coverage over thq
individual’s
coverage.
offer health insmmce
mobile
pref-
&ange jobs more often than
to be geo~aphicfly
Thus,
in the analytical
a job that offers health insurmce
to ad&ess
status.
which may tiect
to be worktig
is lower in jobs whlfi
(3) how does an ti&vidud’s
I“ begh
horn jobs.
of compensation.
we adhessed
coverage
status,
job changes are two
receives from his cuent
Some individu&
the flexibihty
of less mobile individual
b an indivi.dud’s
coverage
or empirically.
of their health insmmce
individuals
by the sorting
~d
studies,
does not account for tidividud
among indvidu~s
themselva
than less mobile
coverage
coverage.
as we~, H this assmption
studies ofjob-lo&
be...the type of worker who is often distissed
are more Wely
individual’s
health insmmce
m indvidud
coverage status.
others for reasons which are independent
which me temporary,
of a
a Edted
of prior job-lock
his health inswmce
either theoretica~y
whi&
represent
papers may not be v&d.
work presented
md
do not fiect
is comterfactual
erences and other unobservable
of job &anges
asstiption
and the frequency
aspect of the problem
This assmption
effect described
h
on job mobility
may affect’ health insmance
that the. offer of. health insurace
camot
of Youth (NLSY).
coverage
labor supply which fiect
that job mobfity
this importmt
Smvey
and mobility.
own labor supply decisions
Employment
this fact impfies
Longitu&nal
coverage.
(1) what factors detertine
(2) how do increases h the price of
ad
health insmance
%ect
the probability
coverage
status;
that he wiH have
coverage?
questions
by constructing
ES-2
a model
of on-the-job
se-ch
.whi&
includes
the value of health insmace
benefit of stayhg
in his cwrent
to receive from his cwrent
given the possible
the mmket.
employer,
individual’s
radt
of potential
is an equation
to insue
wM&
effect of pre-etisting
and medicAy
medical
benefits
expensive.
Wnor
the
he expects
whi& he fight
receive from
the fitidy
demmd
acceptable
for health insmnce
him.
describes
conditions
comp=es
a job offer from the m=ket,
utifity level representing
insmers
individud
benefits whih
the ckcmtances
level of demand for health care will deter him from &mging
which me prolonged
jobs.
on job mobfity
or tempor~
mder
whi& ~
The equation kphes
is strongest for tiesses
ikesses
are tiely
to have a
effect on job mobifity.
The empirical
job &ange.
k
insmance
section
of the paper uses data from the NLSY
the empirical
in a respondent’s
section I employ
measues
region of residence as explmatory
coverage.
tinuity of health
Eded
insmuce
value of accepting
of wages and health inswance
of the model
that the deterrent
sigficat
employed
in the model depend upon an individud’s
and the wi~n~ess
The fist
to the =pected
chooses a resermtion
job offer. Thus, job ch=ges
coverage
fn the model -
job, given the wage and h-lth
combinations
He optkaUy
coverage.
The fist
insm=ce
health insmance
&ding
coverage
@verage
of the supply
instates
hdividuds
the proba~lhty
of health insurace
vmiables for job &anges
is that there is subst~tid
over time.
to estimate
geo~aphic
ad
chages
variation
of a
coverage
in health
in the con-
are more fikely to lose their employs
with high and rising costs of health care tha
in low cost
states.
Second,
incr=ses
have a sigtificmt
in the employer
and negative
when they me in a geogaphic
=planation
obt titing
of ttis
resdt
cost of providing
effect on job mobi~ty.
is that individuals
ins~ance
hdividuds
region with high and increashg
a new job with health fisuance
the costs of health
health insmmce
coverage
coverage to employees wiU
=e
less hkely to &mge
costs of health inswmce.
do not wish to change jobs
coverage
are. ristig,
are relatively
poor.
jobs
A possible
when the prospects
h an environment
for
where
the supply of health insmmce
coverage is fikely
to ~=ge
characteristic,
to be low.
Third,
negatively
I &d
that m
comelated
in~vidud’s
to the probability
propensity
jobs,
-
mobserved
that he wiU receive health inswance
ES-3
coverage from m --
is
ployer,
whi&
provides
strong e~,idence that the job-lock
The analysis impfies that goverment
cd conditions
employees
mmdates
from new health insmance
may resdt
whi&
poficies
in fewer “employers off~g
effect fomd
prohibit
or reqtire
problem,
ES-4
the =clmion
employers
health benefits,
in etisttig
studies is spmiom.
of pre--sting
to ~tend
benefits
and may exacerbate
medito form=
the job-lock
Introduction
b
the U.S. labor ~ket
resdted
=d
in inmeas-
&anges
it is commonplace
to observe
or decreases in health insmace
h health insmmce
coverage
kdividuds
coverage.
experienced
Simple bspection
by individuals
comody
=p erience different types of job ad
health insmance
leave jobs
that offer health insm=ce
for on-
that do not offer health inswance
s-e
job receiving
tenme,
coverage with&awn.
receive health inswmce
Cuently
mverage
theti employer’s
while others work mmy
insmmce
ye=s
coverage
is that the employer
difference
E&ed
i.e, employer
between
i.e., personal
health =re
risk.3
individuals
Two,
cm
expenses
other..workms
plm
= a benefit
be observed
to dange
leave jobs
yeas
b the
tboughout
their
employers
pofities
=e
Uded
this status quo of health insm=ce
insmmce
outcomes.
muse
Some workers
to have thek employer’s
coverage
cr=t es a relationship
h= two tin
job mobfity
coverage
health ksm=ce
health
system
reveal that individuals
offer of health
often md never
m employer.
changes and employment
cme. 2 This relationship
individd’s
insmance
have
of the job chmges
coverage trmsitions.
that do not, wtie
at the same job ody
t bough
wodd
&ages
for ones. that do. Some workers work may
St~ others tight
in the U.S. and a regime in whi~
a
health
in the U.S, most tidividuds’
The sociaUy relevmt
mobfity,
coverage
for whom job
fimges
not covered
may be deterred
pofides
between
social we~ae
in his ‘out
medcd
of pocket’
by health inswace,
change
distribution
to employers
a worker’s
Each rtises a distinct
from a job
employer.1
job
md his price schedule for medicd
concern.
One,
cme =penditm=,
thus creating
by the prospect
1k the U.S. non-elderly addts usuauy obt ah private health timmce
employer. Ustig data Gom the National Me&cd
coverage
are not E&ed
of dependence
status changes,
to =
=
tis~able
that a job
change
though tither thek own or thek spouse7s
Smvey, Fadey and W3ensky (1983) est~ate
Care Exp endit~e
that eighty-fore percent of tidividuds with ,privatehealth hsurance ti the U.S. obtati it though m employer. More
!ecent citatiom of tfi
pe!ce”tage Y=Y between eighty md eighty-five percent.
2Under the gtidehes
of the ConsoEdatcd
separates from an employer the price to b
he wfl either be ~sued
M. coverage at an employment
pohq
b=ed
when a
&mticaU
1994). He mt
emplowent
b=ed health tisurmce
amoss ~erent
i.e., a contrw t that reduces or eat
states of health, on hdividud
1
wc~=e
kdivid”d
y time=..,
md b
obtati health tiwmce
voup premi-
rate. Oth.rtise,
rat% eitha
coverage k less expensive than hdivid”d
paribus,md the cost of treatment without health burance
‘The eflect of hedtb buance,
medical expen~tmes
cease altogether (s.. Ma&im
Ad (COBRA)
or WW have to pay for health tisurance coverage at m tidividual prefim
way he wiH Wely be worse off, bce
coverage, et ceteh
Budget Reconcifiatio.
of that employer’s health ksmance
a matter of months coverage wiHh most cxcs
tiom a new employer k order to contbne
Otibus
coverage is potentially very lmge.
es the &k of _iat
ion k an tidividud>s
was fist analyzed ti Arrow (1974). Job
fight
cause their cument price of medical
effect of employer
E&ed
health insmmce
may reduce
the rate of job mobfity
implication
of changes
considered
reactions
whenever
through
Several
mueller
attempt
(1993),
h health
&anges
(1994),
regdation
of employer
have ternled
about
the
which
below its efficient level, ‘job-lo&’.q
The
for each of these two outcomes
shodd
be
changes are ~kely to cause opposite
support
effect.
and MoAeit
(1993),
studies by Matiian
Holtz-Eakti
Two of the studies, Madrian
the hypothesis;
the hypothesis;
(1993),
and the last study protides
a coflicttig
hypothesis
(1994) ad
two of the studies, Holtz-Eakin
boay of empirical
effect in the U.S. The fact that it has been dlffictit
since the job-lock
Recent
has sud
ambiguous
covwage
Cooper
ad
Buch(1993)
reduces
hfofieit
(1993) and Penrod
support
evidence
(1993),
for the hypothesis.
evidence on the etistence
to show conclusive
httitive
(1994),
and Pemod
that a current e~T~ployer’s offer of health insurance
changes.
Thus, these studies represent
terious,
who have written
this effect,
because re~atory
the job-low
Cooper
to test the hypothesis
do not support
on job mobility
in the labor m~ket
are proposed,
Authors
these two channels.
and VNetta
strongly
coverage
insurance
studies have exafined
the frequency
care to “increase.
of a job-lock
ofjob-lo&
appeal and the casual evidence
is mysof job-lock
is so pervasive.s
I rogue here that existing job-lock
changes
papers have obscured
and changes in health insurance
view that job
mobi~ty
decisions
=e
coverage
the structural
by approaching
made in a static environment
relationship
bet~veen job
thek analysis from the point. of
and with perfect
certainty.
b
mobitity repr=ents another SO”TC.of risk for an fidivid”al, beca”s. an tivolmtary job etit may cause + &&vidu#s
Pi..
Ofm.~cal t,ea~m,nt *Ochange. Hence, there is a risk of variation ~ = ~~~idu~’,
me~c~ exp.n~tm.s
aCrO$s
and employers. In a later chapter of tfis thesis 1 WUmalyze the effect of miversal (nonemploymentstates
~erent
employer bwed) health tismanc e on tidi”idual
welfare, md show that tiversal
coverage k we~are fiproving
ti th
zegard.
4E job mobifity k the mechanism by which titivid”als
(e.g., IOv~Ovic,
may tipede
the sorting process.
h the price of health tisurmce
,educe the tieq”ency
Wed
they are best .mtched
lg79 i JOvanOfic, 1g84; JOV=O~icad ~Ofitt, 1990; Topel 1986; Topel and Wind, 1992), employment
based health tismance
k an kcrease
me sortea- tito the jobs for wfi.~
to a spetic
From the view of m tidividual
a job change may restit
coverage or loss of coverage. These actual -d
of job changes relative to theti frequency when health hsmmce
employer. It & in this sense that employer Ned
health tisurmce
potential costs
my
k perfectly po~table, i.e., not
co~erage is chked
to create
an he fficiency.
‘Most
kmance
of us either how
or know of someone who has delayed o. forgone a job change for fe= of losbg health
co~erage of theh own or theti fatiy
’s medical -p en.es.
2
particdar,
ti=ges
no paper has exafined
presented
in job-lock
(1) pre-tisting
studies,
~ght
medcal
are ofly
meate
coverage
negative
a jOb-10ck ~ect;
coverage
ov=ly
relevat
of that mechtism
M
portable
continuity
emplo~ent
decision
ass~ptions
certtinty.
in that an indvidual’s
that appem
essential element
However,
of the job mobiity
prospects
the job mobfity
problem.
coverage is dynamic,
decision
-d
wtich
coverage ovw
health
md
individuals
deterties
bow
W
m individud’s
md depends on mcertti
One imovation
I re-exatie
value of accepting
md expectations
(as it relates to &mges
insmance
employers.
and changes h health insmmce
outcom=,
or have a
studies have been based on
is static,
the process
itiormation
for obt tifig
and from prospective
of health insumce
Poficy
events.
decision to leave a job today depends on the tradeoff that etists
depend upon &mgtig
is an individual’s
be inv&d.
may be ineffective
in etistbg
the futme value of continued job semch and the futme
These mlc&tions
tisumce
in the employer cost of offting
- the state of the world
and changes in health insumce
It is-d~tic
of health
benefit.
mobifity
with perfect
iUmtrate thati
causal factors wki~
studies, my
acr~ss employers
to matittin
or discussed in the
portabfity
in M of the etisting
to
h W of the etisting
types of potential
between
job &ages
wodd
the basis for job-lock
the may
the caus~
which is emphasized
of the job
rwtrictive
job mobfity
insmmce
have been
health insw=ce
as u
outcomes
mobifity
(2)
which comects
has not been formfim”d
period of time, since they may muse an kcrease
The models
-P1OYW
~d
to above)
statement
whi~
effect on the abifity of individuals
insumce
between
A ford
one type of causal factor -ong
for -g
an extended
two
papers.
conditions,
md job mobihty,
prescription
(Wuded
coverage. 6 This mechtis’m
in health ins~ance
models
the mectism
coverage
Etisting
a curent
of futme ucerttities.
in health insumce
coverage)
this dynatic
of my analysis is that I develop a dynatic
the conclusions
&
in the fut me, both from his curent
studies do not ad&ess
co~erage wti~
job offer.
includes ucertahty
of previous
aspect of
model of job
over wage md health
studies within this more fletible
fi-ework.
The model
studies
have claimed
mobfity
4Tfi
when inswace
mechtism
ksmmce
hcdth
gmerates
two new conclusion
that the demmd
is Med
tisumce
wodd
for health
to m employer,
k essential to the joblo&
coverage s. ktivid.al
reg=tig
insmmce
the tisttig
coverage
j.ob-lo&
studies.
is negatively. rdated
due to.. the effect of the reclusion
story, because tiess
3
to job
of pre-etisting
a job chmge cotid resdt k a &ange h health
not Cccl locked hto a job based on &
pacbge.
.Previow
rcbce
on ti
cment
employa,s
me&cal
conditions.7
special
case b
may increase
condition
I show that the negative
a dyntic
modd
the probabi~ty
mder
of job mobflity;
of a job
which pre-existing
The theoretical
effect of pre-~sting
chmge
it.
pre-tisting
medcal
job-lock
effect; the model impties that my factor
be able to obtain health inswmce
is a
conditions
I derive a statement
new insight into the causes ofjob-fork.
studies is that pre-existtig
I exatine
model
on job mobfity
of the
wiU reduce job mobility.
previous
wotid
in the theoretical
as well. as ditish
conditions
model dso provides
conditions
conditims
me the sole causal factor whi&
whifi
coverage
A key ass~ption
reduces the probability
ca
that =
create a
indvidud
at a different job may also cause a job-lo&
shifts k the market supply of employer
E&ed
health insur~ce
of
effect.
coverage as m additional
somce .of job-lo&.
The source of identification
of pmameter
beca~e
b each case the empirical
between
job changes and health insmace
the =act
structme
of the esttiation
fact that the job-lock
the reduced
of job mobfity
status.
problem.
effects esttited
form models
The stmcturd
model
insmmce
and indicators
specifications
to d=ge
jobs
is negatively
par=eters
the mrgind
the endogeneity
I we
‘k
pael”
represent
estimating
from
of job
correlated
of. the job-lo&
because
issues by =atitig
model
higmghts
—.
the
a fact that
measmes
effect and of the effect
oft he demand
status me co~related
efits.
k
partictim,
with the probabiEty
or with other measmes
.efisting
for health
with wobservables
if the propensity
in
of a
that he win be in a job
of health inswance
effects of these variables
job exit probabifitiw
of these vmiables
data
estimates
=e biased,
of the probability
coverage
away from zero. k
econometric
of the rel.atiomtip
studies have been mderidentified,
of health insurmce
that offers health inswmce
whi&
the identification
is mclear,
in those studies do not reveal.
coverage on job mobfity
individud
studies of job-lock
obscures. the structme
The structmal
model also reveals that existkg
c~vwage
in existing
I clmify
in previous
of healt h insmance
econometric
esttiates
demad,
on job mobfity
studies of job-lo&
then the
wiU be bimed
do not control
for
to ad&ess
this
to job mobfity..
the National
Longitudnd
Swvey
of Youth
(NLSY)
efisttig job-lo& stuties the exclusion of pre-existing health condtions by hsrumce comptiies is adv~ced
as the b-is
of 3ob-lock. Some individ”& have etisttig me&cd conditions for w&& they ~
treatment h the fntwe. Cwrcntly, most hsmers h the U.S. only pay for me$cd
&agnosed alter the stint date of a health burmce
bow that theh pre-etisttig
a cost of mo,tig
ect to receive medicd
treatment of conditions that ~.
pohcy, wtich USU.UYbegtis at the stat of a new job. k&vid”als
conditiom w~ not be covered h a new health tismance pohq,
to a new employer that may deter an tidivid”d
4
from chmg~g
jobs.
so the excl”tion titroduces
endogeneity
a krge
of job
issue. I exploit
s-pie
the fact that the NLSY
of individuals
moblfity.
consistency
ova
The resulting
a rougtiy
data conttins
15 y-
period
pmmnet er esttiat
with the theory of job chages
the entire labor force history for
to control
for uobserved
es show a &mtic
once I control
chmge
for the mobserved
detetinmts
in sign ad
a clem
detertinmts
of job
mobtity.
b aother
health
new resdt,
insmance
coverage,
coverage
of &anges
the gmgrapfic
to employe=
a strong
I discuss
of job mobfity
as being
coverage,
in health insmance
model of job transitions
tibit
mobfity.
model
in health insmance
vaiation
in my empirical
factors
the dynafic
a
import~t
detertin~t
and dthately
ad
transitions
implications
cost of offering
of health tismance
of job mob] fity. I use measmes of
costs across states as exogmous
effect on health insmance
the poficy
points to the employer
in health
coverage,
insmmce
explanatory
variables
coverage.
These cost
on changes in coverage,
and on job
of these results for potential
health tismance
refom
titiatives.
The NLSY
coverage
is tique
tiormation
examination
among etisting
over a long period
of time.
the assmption
a
cwrent
tidividud’s
The NLSY
of previous
employer
data do have cetitin
However,
transitions
studi=
wi~ be offered
that a
steting
the NLSY is m mderut~ed
ova
coverage and job mobi~ty,
since it is the ody
bsmmce
md
for a large sample
employment
histories
the fist
time. The resdts
the youthfbess
of idormation
of health insm-ce
coverage
to generate
ever
of this md-
coverage horn
with cert tinty.
from
somce
health insmance
offer of health insmance
in the futme
Utitations
longitudinal
I use this idormation
of job/hedth-hsm=ce-coverage-status
ysis mntradct
spondents.
data sets h that it includes
of the NLSY re-
for studying
the dynatics
data set that conttins
of individuals
both health
over a long period
of time.
The adyticd
models
fiaework
of job mobfity.
presented
The dyntic
s-d
model
of Mortensen
s-&
model,
by aHowing ucert~ty
cwent
employer
covmage
=plOyer’s
model
md
represents
is m extension
othws.
of shocks
in the cment
to either the cmrent
wage Or health ins~mce
imovation
to etistkg
of the pmtid
eqtifibrim
on-the-job
gener&ation
over the Ievd of compensation
wage of the stmdad
employer’s
a non-trivial
I derive a ustinf
in the futme to affect the ‘reservation
my model to the r=ervation
possibility
(1986)
here dso
sQ&
employer’s
health tisu=ce
cOmpensatiOn
utifity
level’,
model.
wage levd
kcrease
that WM be offered by the
the mdogous
componmt
Thus, the model incorporates
in
the
or the price of health tismmce
POECY. Expected
of=
of the standmd
fiprovements
the res-ation
to the cwrent
utifity level =d
delay job moblfity,
employee’s
while expected
cost of the h-lth
the probabi~ty
of a job &ange.
decfines in the current
inswance
coverage
reduce
employer’s
wage offer or increases
the reservation
utifity
lev~l ad
h the
increase
Previous
Research
Stoppage
of health timmce
out comes
of job
knptication
chmges
coverage
or reduction
when hdth
insmmce
of ttis fact for the frequency
k a series of =ticles
Holtz-E&]n,
(Ma&la,
1993; -d
for health insm=ce
these authors
coverage
is whether
coverage
of job &anges
1994; Butiuder
Pemod,
of health inswmce
ad
poEcies
mobfity
or not employer
are Eded
V~etta,
199*
tm
h=lth
iusumce
Ma&i-
(1994)
possible
to employers.
Cooper
to estkte
by health insmmce
Eded
=e
has been studied in several recmt
1993) there have been att-pts
on job
coverage
and Mo&eit,
The
articles.
199%
the effect of the demnd
status.
The issue adtiessed
coverage
is a deterrent
by
to job
mobfity.s
The conclwions
Expenditme
Her min
pohcy
a
Smvey
employer,
fid
(N~S)
conclusion
directly
hdth
of studies
to =tkate
insmmce
job mobifity
the effect of health
compmy,
=les
=e
25 percent
coverage.
md Vfletta
(1994)
less fikely to chmge
that health insmmce
in Gertiy,
Holtz-E&in
job mobi~ty
no sigtifiat
job-lock
The hypothesis
health insuance
that is proposed
covwage
wM& wderfies
insm=ce
a dti
Pro~m
FinMy,
(1993)
(SIPP),
for single women
a study that compaes
of job-lo&.
obttin
employment
Pmticipation
job mobiity
Dyntics
Pemod
empiricdy?
etisting
studies ofjob-lock
ti these studies, that an increase in m individud’s
employer
&ect
on his job mobfity
begs two, more fmdmentd
of association
coverag$
operate
between a we~=e
and (2) by what mechtisms
The theoretical
fr-eworks
=d
emptid
at best lead to very incomplete
to misleading
sI refer to these =ticl=
no evidence
wi~ have a negative
reasoning
worst they l-d
ad
b
Mofieit
when they receive
reduces job mobi~ty
heads.
though
(PSID)
(1993) fids
effect using the SIPP.
coverage from his cmrmt
hdth
Cooper md
(1993) uses the Panel Smvey of hcome
h the U.S. He fids
insmance
tied
a health insmmce
to their outside coverage,
jobs
Me&cd
on job mobifity.
POUCYnot pmdased
the NMES,
Using the Smvey of bcome
&d
covwage
who have pu~-ed
Mm ustig
but not for stigle men or for men who are household
to esttite
insmance
is 25% lower for those who, h addition
related health tisw~ce
in the U.S. ad
uses the 1987 National
i.e., a health insm=ce
coverage from theti employ=.
that ~ried
BuchueHer
WY.
is that within the group of employees
horn -
insm=ce
of job-lock
amwers.
coUect ivcl y = ‘job-lo&
articles’.
when he receives health
quwtiom:
loss for individu~
C=
we expect
approa&es
demand for
(1) what is the
and employw
this association
to
that have been used h
answers to these two questions,
and at
A formal treatment
qectation
of futme
potential
jobs
mcerttinty.
A determination
employer
often.
of the futme
fire.
slag or tith
compensation
insurance
changes
to employees
fight
=d
considerable
Health insumce
coverage
his company’s
pretiu
may decide
that offer.
an employer
Each
mmket
the probability
hdth
contribution.
to st=t;
in the employmat
places on his cmrent job ad
job
or with a
&mge
to discontinue
that occw
his curent
revenues increase
a demotion.
to employe=
may dedde
as we~ as the &mges
affects the value that an indvidud
of both
rate or a higher employee
who does not offer health benefits
since m individud’s
value of a job involves
A given employer
to include a higher coinsmance
who offers health
his valuations
They may increase when a fire’s
often, evm at the s-e
padage
for a model of job mobihty,
offers determines
they may decfine when revenue
also ~mge
benefits
k
in the ~ket.
ad
is tiportmt
compensation
Wage rates &age
promotion,
~Y
of expectations
of th=e
in general
that he WW leave
that job.
The models
used in etisting
they do not hclude
below,
of individual
which includes a reasonable
ual’s ~pectations
a new job,
=e
an account
studies do not represent
of chmges
that determines
The model
impfies
tisspe-cified
ad
conditions
mobifity
that:
if enacted,
E&ed
to allow employees to
compaties
horn =cluding
the employer
and Methodological
one of may
coverage
or if he wwe to se=~
wodd
conthue
for
Issues
the health
factors
whiti
regtie
compared
medical
health
insurance
may
solutions
to a regtie
to the problem
hsur=ce
coverage
to
from coverage,
They
of national
insmace
and proposals
conditions
measwes.
medicd
cause job
their health
coverage mandates)
fi a regime
cost of offetig
studies .of job-Io&
of pre-etisting
to pufiase
pre- tisting
etist
etisttig
legislative
or touter-productive
or restrict
employees.
Theoretical
model of job. chmges,
(2) the exclusion
(known as cent inuation
to the level whi&
to etiate
used h
and (3) cert tin proposed
mployer
and, by increastig
effect;
health insmmce
are fikely to be either tieffective
cause some employers
because
The model shows that it is an individ-
specifications
p“o~cies is ody
coveragq
a previous
mobi~ty
acctiately,
a dynadc
either at KIS cmrent~ob
a job-lock
insutice
such as proposals
to increase job
coverage,
of expectations.
(1) the empirical
do not identify
health insurance
I d=ive
proc=s
job mobfity.
health inswmce
c-overage though
prohibit
accomt
to be lower in an employer
of job-lock,
=pectations.
in compensation,
from new health
of natiod
this decision
=e
health
coverag+
they
tiely
insmmce
they may
offer to thek
Thmries
of the de-rid
196&
Patiy,
levds
(e.g.,
md
for h+th
1986; Ekfich
Finley
the decision
197% Ffim
kswa~:
and Beder,
and Wllensky,
to chage
and Hech=,
(e.g.,
Lee,
1982; Ma@onald,
of this e=fier resem&
rektionship
hdth
the kt~
fitaatm~
is a coro~~
Structmfly,
tiom.
H~Ith
insmance
ins~~ce
higher pay, better
or because
frequent
ch=ge
a job is se=ond.
ihess,
an intividud
employers
raponsibifi
moves.
when he subscribes
wiU cert tifly
occw
does.g
ance depends
win occu
he has for obttiting
-
because
both because
if he leaves =
of the structud
is feud;
vmiation
An bcrease
his health
a bett=
job (e.g.,
duty);
-y
restit
in a
price is very
pric=
amoss
to be re-evduated
in the price of health insmace
&ange
the nmber
coverage,
rea-
that include
in tismance
me Uely
bm-
for personal
in health insu=ce
the expected
status,
health inswance
or employer
for one that does
that does not offer health insmance
does leave an employer
health
or active dfitmy
that offers healt h“insmace
employer
effect
may fiect
or involmt=y,
A &ange
in
in both direc-
for a M=tyle
risk chmacteristics
pohcy.
rms
because
mobifity,
volmtary
coverage.
leaves an employer
b=ed
may occu
job-lock
coverage.
of a layoff, fing,
geo~aphic
Jovanovic,
that the job-lo&
job mobtity
job responsibilities)
whether
fidividud’s
among other ttigs,
employer
Wustrate
=d
1974),
which is presented
but also job mobiEty
job mobifity
ties, ftiy
to a new health immance
When an individual
upon,
coverage
occw
price of health tisu=ce
if an individud
not, and a decrease
my
Any job ch=ge,
is the norm, and be=use
197~
pictwe
than tkt
wotid
Pady,
Rosen,
for the cment
i.e., they may have a preferenm
in the event of a job &age,
198&
and its effect on health hsurance
more enjoyable
is footlmse,
k m individud’s
fikely to occu
tht
fdy
Rosen,
the bad~oud
relatiomMp
1974
of fringe benefit
and Mortensen,
md job mobifity
Volwtmy
work conditions,
geographiml
Budett
mverage may affect job rnobfity,
job chmges
1990;
reveal a more complm
between health imwmce
hvolutmy
or because
sons (inclutig
coverage
of Me cycle job mobfity
=usation
coverage.
cmpt~
insuance
1963; ,,kow,
the dete-ation
Mit&efl,
197&
wotid
Amow,
1979),
1988) fom
A model of this stmctud
outcome
1993;
Bmdett,
art icles. An ov=view
between
(e.g.,
197% Viscusi,
1983
jobs
coverage
for one
fi his price of insw-
of potential
opporttities
and the cost of health insurace
h
his state.
The job-lock
expects
hypothesis
that a job chage
is that an individual
wodd
cause a
basis of the deterrent is that an individual
ticrease.
wiU be deterred
9One re=on that m employer cm pmchwe health timrmcc
Hence, thae
more cheaply th
of tiurance.
9
chmging
h his price of health insuace
has this =pectation.
action costs create m economy of scale h the puchwe
from
jobs
when he
coverage.
The
must be a mehtism
cm an h&tidud
k tkt
trms-
whid
creates
the bask of the job-lock
that pre-efisting
in fact ofly
effect.
medical
one of mmy
A comon
a formal exposition
somces
implications
utifity is specified
The job mobifity
utifity
as a fine=
of wage ticome
where M* is a latent index of utihty;
and I“ are zero-one
kdi=tors
C represents
and M is a
The mdysis
studies,
me
to job mobfity.l”
is that job mobfity
is presented
~
a static
job offer.
b
is modeled
comparison
kdividud
host
of the
chooses
the
identicd
in that
The ~esdti~
models
indicator
of job-lock
cme utifized
medicd
conditions
the fidividul
focuses
1°1,ho~ that ,ho.h
(2)
-d
a, ~, ad
prospective
=d
alternative
jobs;
1“
jobs offer health insmance
~ me. partieters;
on the sign .of the par-eter
hdicators
u is a r~dom
me included
error
the stat
demnd
for his (th&r)
date of ay
Hence, these indicators
proxy
coverage,
cwrmt
for the level of
metical
new health tisuance
of the demmd
effect of
in C, the cost of changing jobs.
of health inswance
(or hls fdy)
pre-exist
~ in (1), the marginal
of the demand for health insmmce
wife is pregnant,
is that indicators
which wodd
jobs.
<0
of job mobifity.lz
by an individud
to chmge
(1)
w’ md Wa me wages on the cmrent
costs on the utifity of a job change.
medcd
t~c+ ti
if Lf*>O
the cost of changing jobs;
The logic of this specification
timance
based deterrent
job-lo&
– 1“)
of whether or not cmrent
such as whether or not a respondent’s
lzThe
in efistbg
and health insmance.
ifM*
1
{“”.0
“T&
It shows
to the foUowing one:
M=
moving
of this me~atisrn.
costs. 11 The ~tu&e~ ~e
M* = O(W’ – w“)+p(~’
ter~
which is
of this fact below.
decision
net of moving
function
of this mechatism,
of job-lock
of a health insmance
job offer with the uti~ty of an altaative
that offers the most
cov=age;
exposition
which are the sole somce
and empirical
process.
of a curent
me stiar
conditions,
ay
bond among the articles which have mdyzed
as a static decision
job
articles ofit
My model includes
potential
I discuss the theoretical
utfity
Job-1ock
this expectation.
conditions
pofiq,
for health insmmce
–
were
coverage
t. tl,, price of medicd care are another potential caue ofjob-lock.
model was fist wed to stndy fitige benefits and job mobflity by MitcheU (1983).
~er
~ode,,
mainly fi ~heti specification .f theter& w“ and I“, “the emected
offers, respectively.
ch=acterktics.
kother
measwes of h-an
One strateW is to pre{~ct w“ md I= ubg
specification
of (l), used by Ma&im
(1994), otits
capital characteristics, such as years of ed”cstion
econometric specficat iom of (1) and (2) are discussed below.
10
tio~mation
m=ket
on tidividud
these term
wage ~d
he~th
h-an
capital
altogether, but ticl”des
and yems of employmmt
expedience. Previous
proxy
for the cost of changhg
existing
medicd
inte~reted
= evidence
However,
hdth
conditions
ud’s
mpected
mce
merficd
to theory,
coverage.
tidicators
uti~ty
demand
C,
mderlykg
the supply ofhedth
conditions
n
individual’s
vidual,
wodd
by C).
suggests
be usefd
pting
coverage
coverage,
kcrease
1$T~
-y
job
because
be Wowed
both
in an tidivid1976).13
to their demand
Hence,
for health insmimwmce
demand
to affect ~, the mmgtid
C and @ my
Stated differently,
be affected
an incr-e
the exclusion
by an
h m tidividud’s
coverage though
(given by ~), and it 4s0. may indirectly
ofpre-etistkg
deer-se
medical
to separate out these two effects using indicators
of health inswance
supply, if aogenous
y. I argue, below, that measues
across states are exogenous
mobfity
to have theh fist
tid
irr theh dem=d
of
to an indi-
of the geo~aphic
to hdviduals
inme=e
than deter it.
ti the near future.
“h a
ad
kdividud’s
For &ample,
H neither
vmiation
are detertinmts
of
a yomg
couple is
receives health immace
coverage - nmely,
that m h~vid”al>s
is wfig
m~gtid
jobs.
other wiables
utfity of medicd
to pay more for a mmgtid
constmt.
11
Shce
theti ficreased
to when he is healthy, then the shadow price of health &urance
for the si& thm for the healthy, holtig
for medicd
may begin loobing for a new job that offers health insmance
for health insmance
k sick thm when he is healthy. E an htivid”d
demmd
suppose
of thm
jobs offer lower wages thm their cment
~e~t~on the ~s-ption
when he k si& comp~ed
that m
rather
a job, one or the oth-
even if alternative
~~~
Sti@tz,
coverage.
alone.
raises the possibiEty
induce
though
coverage
demmd
for identifying
hsw~ce
issue – the same health
shotid
b
alternative job offer,
is kcrewing
ad
the &ange
supply.
TMs argment
services
Rothstild
according
=
increase his demand for health insmam
that determinants
in the cost of health imwace
health inswance
jobs,
mre.
It is bpossible
pre-
value for y is
represents
in health
coverage
coverage offered to him though
health insumce
This =g@ent
197*
coverage,
utifity of su&
insu=ce
(represented
of health bsm=ce
for medical
esthated
were to accept
utifity of that change
across individuals
care may directly
its effect on the m=ginal
if m individud
of the identification
demand
a negative
exclude
hypothesis.
(e.g., Arrow,
insmance
articles
POHCY wotid
frr (1) the term (1” – 1“)
the cost of changing
in health
for medicd
utifity
WY
This is the cru
that inter
kdividurd’s
idmtified.
e~enditwes
~ shodd
of a ch=ge
of the job-lo&
the ~ginal
that the mmgind
a new health insmace
k job-lock
that would occu
~ represents
predicts
accor&ng
in support
coverage
so the parmeter
because
from coverage.
y is not theoretica~y
insmance
Theory
jobs,
the cause of the
dem~d
for &l&a
treatment is higher when he
mount
of medicd treatment
coverage WN *O be figher
– shodd
not have any mrginal
probabfity
effect on the employer
of a job etit wodd
increase
The decision to change jobs cuot
such as the framework
dt-ative
job
off=
to the expected
cmrent job.
utifity
(2).
the mpected
he wotid
receive
h a static frmework
ez post compensation,
compensation.
h a dynatic
modd
wae
distributed,
individual
utifity
fimework
expected
cmrent
compensation.
Some compensation
may represent
=
Alternatively,
e~ected
the value of staying at a cuent
in compensation
job -d
have the opposite
The previous
two puagraphs
fiect
may expect
conditions
tisttig
increase the probability
imply that (1) is tisspecified
ez post wage -d
studies of job-lock,
so the question
ague,
below,
measmes
miables
Empirical
whose focus
employer,
of hman
capital
to predict
the m=n
v=iation
ae
to be fied
or
decreases
expected
in
wbid
lower
incre=es
has been
coverage
an individud’s
in health insm-ce
importmt
detertiats
of the distribution
coverage
is
.12
insm-ce
what -e
offers?
the effect
treated
offers.
The
the detertinmts
That
question
of pre-tisting
coverage
health insmWce
health
or wage rate.
as an exogenous
coverage
is
b
variale,
never mises.
I
costs at the state level as we~ as other
of health
of health insu~ce
Issues and Results
w’ and 1“ represent
while the appropriate
ez post health
of health insu~ce
health inswace
of what factors ifiuence
that geo~aphic
randody
ti compensation
in that the term
health insmance
and not the det ertinats
cmrent
=d
effect.
the ~pected
those studies an individud’s
since they
promotion,
of a job ch=ge;
one, because it suggests the foUowing questititi
on job mobfity
ec ante
at his cuent
expected
decreasu
ez post wage ad
not asked h
decision,
an impendng
Expected
wodd
expected
job, i.e.,
i.e.,
may =pect
cases represent
specification
of m individual’s
job
stay at his
receive,
b,e wexpected
an individual
offer from the cmrent
tiportmt
m
the dtemative
at his cmrent
the mobifity
BY
of the ez ante compensation
difference is=
of accepting
level of compensation
components
include
fraework,
offer ad
rective
thefi
are not eqtident.
Both
increase in compensation.
horn
alternative
chmges
For -ample,
an individud
=gment).
that he cmently
An indlvidud’s
layed off, or may know that he wi~ be forced to qtit.
by the job-lock
receive
that he wotid
the two quantities
but others may be expected.
coverage,
assesses. the utfity
he to reject the
changes h compensation
for myriad reasons.
insmance
in a static theoretical
he wodd
to the compensation
affect the value of staying at one’s current job.
job may ch~ge
a
repraented
the compensation
is equivalent
h a dyntic
When
of health
as is predicted
be accurately
in (1) ad
he compares
(not dec~ne,
supply
insmmce..
status.
I use these
offers h the labor m~ket.
Although
coti
etisting
job-lock
three previously
ace
coverage
Buchanan,
ploymmt
md
studia
-tabfished
dfiarent
facts.
conclusions
with regmd
individuals
with employer
One,
me less Ekely to leave their jobs thm
and Leibowitz,
=d
continuity
McBride,
wage income
r-rh
1990).
1992).
of private
Three,
(Kle—,
Two,
there is a positive
health insmace
there is a positive
1992; Kronirk,
coverage
to the labor
tha
force ad
me of high SW
previom
ity that an individud
that ~
individud
cmdd independently
job
&mges
coverage.
has a health insmace
jobs
poficy
we wodd
assign health inswace
const~t
d
coverage is not independently
opportdties.
A more redstic
vary as we VWY an exogenous
inferences
for health
strategy
detertinmt
as to how job mobfity
nrined propensities
assigned.
other
Mther,
wodd
detertin=ts
of hdth
attatient
health
insmance
~e of low skin.
ad
the probabil-
the probability
ksua.ce
coverage
wherein we
ad
observe
of hdth
insurmce
it is m outcome
coverage,
of =
coverage.
thek
inswace
and
individud’s
how individual
mobfity
We codd
with different ~ogenously
This is the proposed
md
to offer health
between
of individuals
of health insmace
covaage.
U&ed
employer
be to extine
diffws across individuals
insmance
benefits
to have m experiment,
to a group
across hdividuds
1988; Swartz
who have a strong
task to show that health
coverage
of em-
These three facts te~ m
to the labor force =d
w-t
(Kleru,
mntintity
health insurance
a negative relationship
iddy
of thm
health insur-
and Schm,
1990).
from his cment
it is a greater
To show caustity
between
me more Ekely to have elnployer
studies have show
H&ed
long lasting, =e more Wely
k reafity we cmdd never control for the det ertinants
rhoices ad
mde
descriptive
while holding
health inswmce
patterns
=e
McBride,
who have a we& attachent
wi~ &age
reduces job mobfity.
wd
between
d
such coverage
(Monheit
those that do not, and that individuals
coverage thm those individuals
Mthough
coverage
1991; Swmtz
without
relationship
correlation
that good jobs, i.e., those that pay tigh wages ad
insm~ce’
individuals
to job-lo&,
then
deter-
strategy
in the etisting
Although
au the authors
studies of job-lork.
The authors of these studies atbate
begin their analysis ushg
~que
the sbple
ecOnOmetric specification.
the fo~owing
two dtemative
the probab~ty
theoretical
of a job &mge.
fr-ework
COOper and MO~eit
job regime equatiom:
13
in (1) =d
(lgg3)
estimate
(2), earh author
(1) directly.
chooses a
They spec~y
where
XWj me person
pm~eters;
md
are stimly
i’s
eui is a norma~y
that
distributed
deterdne
his
error term.
~d
Mofi&t
wage as the dependent
or not ~
individual
health inswance
the predicted
est@ate
receives health insm~ce
equation.
at Ms cmrent
Using the estfiates
values of I? md
w: for every hdividual
the residuals from the insmance
of
for I;
~ me statistically
significant,
gtin health inswmce
th~
those who are predicted
=e
consistmt
FinWy,
and both -e
coverage
employer)
indicator
of whethw
~iable
to lose coverage
model
]? ad
ad
the tem
the term (w; - w;)
(1) ~d
for job mobitity.
ti,~,
with
(2) by addng
Thek
esthates
Thus, those who the model predicts
(low 1’ – Ia ) are more Ekely to ch~ge
jobs
(tigh
that. having
h the
They represent
they esttite
negative.
after chmging
\vith the hypothesis
(cment
job as the dependent
regression,
to the right hand side of a Probit
term
zero-one
i k their smple.
these two residud
resdts
of the equation
of 6W and 61 they create the variables
from the \vage regression.
jobs
bw is vector
ratq
using the observed
and the observed
(w: - ti~), the residuak
wotid
wage
The components
the two regime equations
variable in the wage equation,
with (I: - f?),
of a -d
offered
defied.
Cooper
(I: - I?)
characteristics
1“ - J“). _Using ttis interpretation
employer
Htied
health ksmance
their
coverage
reduces job mobifity.
However,
their results may be just a spmious
Ekely to be unobserved
chmacteristics
ability that an individud
that m fidividual
cal mgments
coverage.
outcome
of individuals
or employers
whi&
wi~ be in a job that offers health insmace
wi~ ch~ge
jobs.
that large employers
For ample,
tisspecification.
deterfine
coverage
we know fronl previous
me more Ukely th~
smd
employers
FtiaHy, we kow
that mobfity
which tidividuds
employers
ad
who have Amacteristics
individuals)
erage in terms of mobiEty
~U,
Iazge
.mploy.,,
are sdected
a ~,,.t,,
than do sd
hc,mf,v.
ticreases the awstrat
empiri-
compared
to sma~
at sd
employ-
tha
of a selection
proc-s
in
but known to both
of productivity
and below av-
E such a selection
process
employers to select less footloose, i.e., less mobile,
iv. cost of both health fiuanc
e md pension benefits, -d
employers k offering both types of benefits. The observed correlation
between lmge employers and figh wage rates hbe ea~y expltied
ad
to offer health insmance
to the econometricia,
into jobs with large employers.14
Imge employers have a cost advantige over sti
that the correlation ._ot
outcomes
which make them above average in term
ha,.
workers, b eca~e job mobfity
(wobserved
~e
are
both the prob-
theoretical
rates me. lower at lmge employers
ers. It may be true that the latter two of these correlations
There
md the.probability
We dso know that average wage rates are higher at large employers
employers.
14 After
of model
bce.n w.U doc-ented,
but is more WCdt
to expIati.
The f-t
is it se~ evidence that it is the outcome of mob served chmactetist its.
14
-.
operates
in the labor -ket,
be more
fikely than m tidividud
health
insmance
More
between
then an individud
and less ~ely
fomfly,
without
there may be uobserved
< 0, then esti-tes
insmmm
coverage
Mati]an
those characteristics
than same to &age
u, the error term in the mobfity
ff E(v er)
with those mobserved
to be observed
wifl
in a job that offers
jobs.ls
attributes
equation,
selected Aar=teristics
of individuals
md er, the error tem
of E WU be negative
tbt
create
a mmektion
in the inswmce
even if there is no causal M
equation.
from health
to job mob]~ty.
(1994)
does not esttite
(1) ~rectly.
Her econometric
specification
of (1) =d
(2) is
as fOHOws:
M; = blw; + bZI: + b3Ci + b4(Ci I:) t u;
where the measwes
size, and =
of Ci k her study are whether or not a respondent’s
indicator
poHcy not fified
equal to one if a respondent
to his cmrent
ksm=ce
from theh cmrent
addition,
have a high demad
not.
Hence,
she ~ects
Her restits
ployer E&ed
cofim
the cost of ch~ging
in~viduals
who receive health
jobs should be Mgher for those who, in
health insmmce
coverage than for those who do
of the Cooper
=d
Mofieit
study, that tidividuds
=e less Hkely to &ange jobs than indviduds
and Mo&eit
another health insw=ce
.of b~ and bq to be negative.
the main.resdt
paper, this resdt
to job mob fity, because
tests for job-lo~
demand
hteractions
that song
fdy
with em-
without such coverage.18
does not nece.ssmfly imply a causal relation-
health insmance
coverage is endogmous
indicators
me derived horn her esttiates
(pregnancy,
with health insm=ce
no outside
of bs and b4, the @efficient.
health inswmce,
and fady
of health
size) ad
employers may be more &ely
than others to both offer health bsmmce
Employees of those compaties ti
coverage md be prefemed by workers.
be more Wely thm the average employee to have health tisumce
of & == ,tatkticaUY si=ficmt
may ~= ~ob,er;ab~e
change and the probabfity
for a fd
some
coverage, and
thm sam. to change jobs.
16~er ~t~ate$
1,Therc
thek
coverage.
%.An ~nalogo”, ~gum=nt may be ~ade .n the bwis of employcz rather than tidividual ~=acteristics.
1.ss ~cly
to the
equational’
Ma&,m’s
insu=ce
She proposes
for employer ~ed
her esthates
ship from health insm=ce
job mobi~ty
employer
health insm~ce
Just as in the Cooper
employer.
wu not covered uder
wife was pregnmt,
exposition
&mact=fitics
at the 5Y. level and negative ti ~ of her sPec~catiOn,.
of b&viduds
that a job w~ be one that hcludes
of this petit.
15
or jobs that detertie
hcdth bmmce
both” the prObabfity Of a jOb
coverage. See previom p~a~epb
Abost
d
of these coefficients
a negative ad
sigtific=t
coefficient
(table 4) md a negative
pregnacy
(table
a statistically
sho~v that a spouse’s
that it shodd
Mso,
example,
d
to the interaction
coeffiii{rit
the stmdmd
tkee
errors
measures
to have a spouse that is pregnant
wotid
bs wodd
be negative
produce
dso be negative,
ad
correlation
E(1 C) >0
individuals
estfiates
of a job
coverage ad
appears
concluding
chage,
that hMso,
when
to the job mobility
who me in ustable
between
C ad
her
theory
correlation
may cause a negative
of b3 even if additional
fatifies.
employmmt
For
me more Mely
situations.
TMs
then esttiates
md estimtes
of
of b* wodd
became
fided)
outside health inswance
of outside
between
health insurmce
ti@ed
coverage.
health insmance
C and u. 19 Hence,
~
coverage
~aTP2)
coverage
unobserved
C, which agti
wodd
were not a detetinant
+ V~763)
= (–.039+
correlation
would show up in (1)
‘a For ,=MPIC, in table 3, specification 3, she reports a p-value of .017 for ‘Job-Lock, Test 1’. ~e
shows a norma~y distributed test statktic, z = (B2+b3)/
who
puts you in a
This positive
coverage
infividual~s
u md
health
U YOUare & mob~e person it may
you know that YOU &gh job moblfity
for lostig your employer
correlation
(not -ployer
It may be true that individuals
are more fikely to pwchase
md pmchase
between
equation.
situations
u. E E(C u) <0
who do not have outside
who have lmge
outside coverage,
between risk of job &ange
mobfity
the probabihty
by below average job mobility.
average risk category
as a negative
erroneously
size
ii this case.
average job mobihty
seine to pmchase
higher tha
fdy
many cases Matii=
who =e in stable employment
thm individuals
a negative
than those who are &aracterized
mde
h
do show
when in fact it does not.18
of C are endogenous
TMs point is equally true for individuals
have higher thm
insignificant.
test statistic
coverage =d
of health insmmce
even if C were not a muse of lower job mobtity,
shce
coverage
~e
increases
it is quite possible that individuals
inswance
to the interaction
her restits
reduce that probability,
of Ma&l~’s
tendency
of health inswmce
of her test statistics,
sigtificmt
pre~ancy
at the 5% levd;
insigtificmt
5), but au other coefficients
evidence- supports
predicts
ititisticdly
md si~fic=t
to have tiscalctiated
rwdts
tie
prochtity
produce
for job
negative
of job mobfity.
COrItc! cdC~+t!On
.21i)/~.1075’
+’.13391 =
1,00, or a FVAU. of .159.
19The ~=a~mc of c h tti, .,se is ‘no addtional
mobfity
costs. kdivid”als
who own additiond
lower refimce on theti employer hked
do not have additiond
health insurance coverage’. It is a shghtly Cofiwhg
m.asme of
(not employer finked) health tisurance coverage ~e said to have a
coverage. Ha..,
mob flity costs, C, me lower for t~s ~oup than for those who
coverage. U, h this case, C is defied” as one foz those &&vid”&s
who do not have tititional
coverage and zero otherwise, then a p otitive value of C “represents the group with the &gher mobs ty costs.
16
Est*tes
of b, wodd
employer
dso
Haled coverage
have employer
Fmfly,
~ed
=e
less fikdy
of chil~m
a detemrimt
they are in a stable job before
C, whi~
supply -d
(1993) attmpts
levd
comelation
itse~.
&mteristics
comelated
hdividuds
who have
who do not
on a
then there will be a negative
correlation
of tilfien.
is increashg
in =
h this case, = h the
in the nmber
intividud’s
of cM&a
age. Ag*,
of measmes
(1) by t we-stage-least
-ogenous
exogenous
equation,
measwe
esttitw
of health insmace
squares, instrwenting
for m individud’s
set of chmacteristics.zo
measme
of -
tidividud’s
since the two-st age method
predicted
Pemod’s
demmd
clfi
for hdth
etinates
for
is
insm-
the uobserved
and job mobfity.
in the job
stage equation
and health
inswmce
creating
=e
mobfity
predictors
coverage
h twn this fact impfies that the instrwent
equation,
wait ~tfi
He creates an instru-t
used in the fist
term in the job mobtity
costs, because the risk of a
H bdividuab
the issue of the endogeneity
sti~ is endogenous
the instrument
of moving
due to the effect of mob served vmiables.
demad.
=
the demmd
his instrment
Thus,
thm
to have a fdy.
demnd
He estinlates
b~ed
in a job mobfity
between
However,
ogenom
equation.
represents
measme
size me increasing
negative
expendtwes
that the hstruent
coverage
coverage
as the nmb=
hsmace
f~y
of health insmance
of medicd
=ce
shce
to addess
to a job mobifity
the measmes
to stat
in this case is measmed
bq may be spwiously
d-d
is an endogenous
a fady
two, E(1 C) is positivq
Pemod
outside
&oosing
previous
of b~ =d
to pmchase
decision
u ad
both insmace
in this case; that is, if individuals
of m ti~vidud’s
between
nd
if E(I C) >0
coverage.
the umber
job efit maybe
be negative
a sptious
equation,
W of the ex-
of health hsmmce
status wi~
wiU be negatively
negative
stice
in theory,
cov=age
be positively
correlated
with the error
between
the index and
correlation
job mobifity.
The endogeneity
predictor
si@ficmt
,,pemod,,
of job
issue may be moot,
mobfity.
md negative
Nonetheless,
estfite
~ethod k ~o,.
no medical eW.n&tue
as the instrment
consistent
tior~tion.
with the other
of ~, the coefficient
COmphcated than the stmdmd
Hk ‘b”str=ent’
w~ch he b eEeves me detcrfiants
re~ession along with Mor-tion
for each h~vidud
b &
.XP enditw=
expend tues.
on tidvidud
&.racteristics
did obttin
a
coverage in (1).
k constructed ushg data from hhc
1982 ). Ustig medcal expen&tme Normtion
on a set of =ogenous
He uses the wtkted
tidividud
ch=actertitics
coefficients horn the exp enditme
horn the SIPP to predict a level of m.tied
SfPP sample. Ttis kdex k Ms kstrment
17
Pemod
2SLS method, sfice the SWP data that he use. conta~
for medi.d exp en&tw-
of mc~cd
but was not a sigfic=t
studies,
of health insm~ce
National Me&cd C=. Ut%zation ma Expen&tme Sw.ey (~CUES,
k the NMCUES he rms . regression of total me&cd
was negative,
k the job mobfity
even~tues
probit equation.
ButiueUer
ad
VMetta
( 1994) emphasize
es tfiat es of job mob.ifit y, ,Their econometric
where
Pensions
employeq
is a d-y
Tenure:
is a mmswe
and If is the observed
in put
arWe
theoretimHy
pensions
measmes
theory
of ch-ging
select a job
ad
cost factors.
cost factors
produce
omly,
md
interactions
of the specification
somces
(1994) ad
mobfity
ben&ts
(1993).
w; is the observed
ti the mobfity
wage rate;
equation,
because
job
tenme
of job mobifity.
ad
job
theory
tenure me related
dso
suggests
of pension
These
less fikely tha
retail
tkt
tQ the cost
uobs.ermble
coverage,
job tenme,
individuals
win select
the average individud
higher thm
average moving
to
costs.
wi~ have below average tenme and a higher
offers pension
~s
fin
outmmes
benefits.
since d
(1994)
is sifiar
inter-t
have some
size. Th-e
Hence, in BucbueUer
though
fioices.
in methodology
the wttiates
18
cost of
sh~ed
the meu
Given
cov-
md Vdetta’s
dscussed
of Pensionc.
previOther
on the right h=d
side
these many potenti~
they obttin.
is ti compming
k
benefits
in employer
of the variables
of hdvidud
the esttiates
in the U.S. to that in Gamny.
pension
with mob served det ertiants
are possible,
The employer
that a job offers health inswace
with u &ectly,
its correlation
to fiterpret
cost of offertig
the probability
comdated
represent
in this specification.
sets of costs ~e decremtig
between
the study of Holtz-Eakin
Pemod
more perverse
both
wobservables
above
individuals
and the employer
that a job
though
of bias it is ~ctit
FinWy,
is eva
1’ wi~ be negatively
between
those jobs
his curent
offered as part of compensation.
may be footloose.
They ~e
tith
the value of a job is determined
may be determinants
costs.
one:
They also rogue that cmrent
coverage
in theti
of a job change.
a correlation
indirectly,
jobs.
respectively,
because
For ~~ple,
erage and the probability
specification
employers
imply that footloose
cov~age)
be a predictor
of a job,
which offers a pension,
employe~
be included
some individuals
The effect of uobservables
comon
should
md job tenwe
i has a pension
thek model
insmmce
that pension
average moving
insurance
k
mverage
to the fo~owing
individud
a cost of chmging
For exmple,
retail below
health
heflth
coverage.
does predict
average probablfity
offerbg
for whether
of a job, and shodd
Mso, their preferences
thm
th~
represent
of tidividuab
job mobifity.
jbbs whi&
(other
jobs and the dmabi~ty
chmacteristics
is stim
of i’s tenme with his cwrent
that pension
the dmabihty
Although
model
health insuance..tidicator.
by fringe benefits
The authors
ad
variable
the roles of pension
to the studies of Ma&ia
the effect of health km~ce
for the U.S. he tids
on job
no evidence
of
job-lock.
A miticd
we&ess
of the observed
eq-tion.
h the empirical
health insm=ce
I have described
detertinmts
co&cient
of the health
expect ation of ez wst hdth
equation.
in the emphid
the expectd
Thus, expected
section below.
The difference
in coverage
is theh inclusion
ad
job
mobiity
or employers
-y
may cause -ttiates
be
of the
in the mobifity
equation
model specifimtion
shmdd
(not the observed health insmmce
empirimfly
include
as wefl - I provide support
coverage is the mpected
chmge
the
h the
for the cl&
coverage
in health insmwce
from the econometric
to a-
mverage)
between the observed health insmance
have been otitted
lea~
-d
coverage.
specifications
used in
studies.
This sowce
(l),
coverage
can be supported
job-lock
bi~ed.
insuance
d-y
that the comect
This clfi
chmges
mticles
eristics of individuals
md job mobfity,
coverage
Recaff
ez post health insurmce
job-lo~
to be negative md biased downwmds.
hsmace
problem
~=act
coverage
covwage
of Asting
on the right hand side of the -ptilcd
how mobservable
to health inswmce
other more fudmental
tisting
indicator
of both health insumce
The inclusion
mobfity
specifications
of model Asspe@cation
E =pected
&ages
me correlated
is fikely to cause the esttites
in the cuent
employer’s
with either (w’ –w”),
compensation
of pameters
offer, whi&
(lC - 1“), or C then estbtes
in (1) to be
me ofitted
of the pmmeters
from
a, ~, or
y wiff be biased.
An indvidud’s
k
MS cment
insuace
state of health maybe
employer’s
coverage.
coverage
fiages
shodd
h
The positive
&eady
compensation.
~ected
a negative
included
2I ~
plm
b
~ay
betwe~.
.auceI
h order to -cIude
due to fiess
rep ontibfities
kdividuaf
fi,
insmace
coverage
between
itiess
coverage of a figh
or pregnmcy may fid
have b =n reduced.
the demand
correlation
eithw reductions
for health
for health
between
insm~ce
ikess
and
employer’s
foUowing a serious ikess .21 A negative
comelation
and the de-d
u in (1), shce
,mpIOy..
ad
&mges
in his cmrent
&anges in compensation
,, .
.
...
.U,..t
of titme
offws and MS demd
may also be a negative
benefits
C ad
of both his expectations
may expect
in compensation
in C arid mpecfed
employer
There
of his bedth
rh=ges
correction
health
correlation
be clem.
wage offer or restriction
between
wage ad
a detefinant
hed th &wmce
cost employee.
for ksmmce
h~cators
me otitted
coverage
of. hedt$-
hsw~ce
horn (l).
$. ,
K E(C
b emefitsPlm or switch to.
An employee who t~es
dem.d
me
u) < 0 then
an extended leave of absence
Both outcomes w~ have a negative effect on her expectations
19
cause
uc w health ~~ce
that her pre-leave opportunities for promotion have been mtdd
O*,,.
wodd
or that her
of fitue
wage
esttiates
of y tin
be biased downw=d.
Mso, chmact.eristics
cmrent
hedt h insmace
health insmance.
c=e
of the state ti whi& m employer
industry
coverage
mends
health inswmce.
Hence,
coverage
offer ad
his expectations
As I =gue
in the next
cause the employer
in regtitions
by employers.
tismmce
offer.
characteristics
state to state.
offaed
coverage
tiect
(although
health insmmce
coverage
ad
expected
coverage
h~lth
and health
to va~
state level sho&s
of dnges
WW be corrdated
tismance
born
to the cost of
coverage
k theti employer’s
with u ad
of the bias depends. on the relationship
futwe
employer’s
the level of health insmmce
base theti expectations
the &ection
both an individud’s
h his curent
tismance
produce
offer on these trends then in (1) I’
tiect
state level ~egtiations
cost of health
trends h state level factors
-y
of &anges
section
and characteristics
H individuals
@ wi~ be biased
is located
health
estkates
between
of
cuent
coverage).
Data Issues
Previous
Topel,
stu&es
198&
Heckma,
mat=
of job mobflity
Altonji
1982;
caused
md
G6nfi,
by uobserved
employment
components
of job mobfity
of wages =d
the @ect
to addess
coverage
Estimation
dem-d
my
for medical
in ea&
study
empirical
data conttiting
md
significantly
Empirical
studies of job mobtity
md
wages, titing
compensation
of these
mtegories.
of health
insuace
issues.
longitudinal
between
coverage
insmace
ad
20
and transitory
estkates
do~
~ models
not accout
and health insmance
coverage
for
need
of health insm=ce
job mobfity,
health insmace
tidividud
coverage
itiormation
of job changes,
because
No tisttig
covhealth
tior~tion
job mobfity
wodd
avtilable
include,
for
pretims,
set of data conttins
of cmrently
job mobihty
ad
health tim=ce
complete
data imply
that
to addess
one
studies of job-Io&
use
wiU be uable
that none of the etistkg
.ovwage
esti-
for wages.
The fidtations
StiU, it is smprising
health
offers.
strategy
of wages ad
Ievd data set. Relevant
of mr individud’s
permaent
malysis
Ffirm md
studies use records
They show that parameter
when the esttiation
1986;
198fi
bias of par-eter
These
for’ mobservd
as weU as to m aalysis
individual
cme,
the possible
issues; these issues apply to a
equa~y
and Fmber,
19.93; FUm,
chmacteristics.
to ‘control
r~pectively.
1992; Abr&m
WiUams,
to ehnate
and employer
of the fu~ set of rdationships
the history
empirical
or more
change
Ward,
is a subtitut e form of compensation
m ext emive
ifiomtion
Altonji
attempted
~d..wages,
these same empirid
tismance
example,
kve
individual
of mobservables.
job mobfity
198fi
and wage Mstori=
job mobifity
erage ad
reqtires
ShAotko,
1988)
of kdividual
and wages ( Topel ad
or discuss
the imp~catimr
of
I
longitudinal
data for thek ~yses,
theoretid
since the dynafics
1993), the PSID (Holtz-Etiln,
The SIPP =d
labor uket
import~t
contkued
evidence
for more
another 21 tilhon
md PS~
i~ormation.zs
mperience
&agm
in. the U.S.
k ~
to kow
it is those wderlytig
hdth
Swatz
insmmce
(1994)
mverage
l~ting
health
coverage
1993
199% Pemod,
Matilan,
1994).
used souce
of longitudnd
coverage
cdctiates
between
hsm-ce
that 37 ti~on
over time me a
aericms
eqe-
1 md 11 months sometime
dwing
coverage for at least a year. While the
longitudinal
can measme
hedt h hsmance
the causes of &anges
causes whid
statw
health insmmce
the prevalence
coverage
Of ~dividuds
whO
over time.
tidividud’s
in a
are deter~
codd
not be used to ~dyze
coverage over tke.24
bdividurd’s
As I awe
health imuace
ts of job mobifity
coverage,
rather thm
inswmce
it se~.
Data wbirh include records of tidividuds’
job mobiEty.
h those studies the authors
22pm=~ ~ormation
.ppro*at
& the SPP
work historiw
at tapt
for job changes usiug a variety of econometric
CXtCn& 36 monik,
to control
tetiques.
wtie
have been essential in past studies of
for mob served individual
25 ~
each CaSe ~Ob
k the NMES Mor=t
ion &
served
prochviti~
factors
were
coUect.d for a Period of
ely 15 month.
~,T~e SIPp ~d the psD
one-ttie
and is a mu~,
data sets, none conttin
ti health insmace
it is kportmt
coverage
25 yems,
b
are without
we pmel
of &ages
Valetta,
md MoAeit,
reason, the data sets used h prior studies of job-lo&
the detetinants
stice
(Cooper
ad
22 The P sID is a long panel data set. The PS~
Th~~ none of the efis ting ~dyses
For the s=e
below
thti
(Bu&ueHer
data
rience a spe~ without health insmmce
SIPP, NMES,
the NMES
suggests that &-ges
fact for ti&viduals
=d
me the SWP
both short paels.
and demographic
Etistkg
studies
1993), md
the NMES =e
in 1968, h=
a ye=,
coverage is the central
issue k those studies.
The data sets used k etistiug
begs
of health insu=ce
.oUect,d
supplemmtazy questionntie.
However, ody
the health tisumce
h~dth &urmc.
for a lon~tndmd
Wor_tion
24T0 .mpha,~e
the obtiou,,
mdysis
ody for . skgle Petit k the
The N MES coUect ed hedt h tisumce
at the fist
whether or not the five observations on health bmmce
data sow..
coverage Mo.m.tio.
ion at d five hterview dat es.
date was wed by M.*,
It is questionable
coverage taken at rougHy thee month titervds h a Wa
of health &u~ce
note that tithout
kterview
Wmt
with a
coverage. So fa tti
Ion@tu&d
&sue k moot.
health tismmce covmage data such m malYsb wotid
be bereft of a dependent vwiable.
2bTope~ ad
W=d (, ~g~) provide . detafled &$-&on
of the hportance
21
of lomgit.~d
work &tory
tiOrmatiOn
fomd
to be corrdated
Job-1ock
studies
case of the PS~
of health fiswance
tr-itiom
with relewt
have not contro~ed
it has been possible.
effects
explanatory
for uobserved
c=
ody
in a data smple.
for ntu&es of”job mob~ty.
22
such as job
deterd~ts
The job mobfity
on job mobfity
for each indvidud
vmiables,
Hteratwe
tenme
of job mobitity,
impfi-
be obt tined ustig
tht
=d
wage rates.
although
consistent
h the
esttites
several observations
of job
A Job
Se=ch
Model
with
Insmance
k this section I develop a job sew&
insmace
coverage
individual
md the distributions
the ~ket,
&oice
a wage.
By including
decision =d
&=ges
a dynatic
specification
h =
individud’s
deud
in Bmdett
-rket
of the job ~tting
deterfine
=
of health insumce
cons~ption
component.
The
constr&t.
to this problem
prims of he~th
1, ad
tismace
decision
An individud’s
the distribution
is represented
wage b the
utifity level which reflects the
detertie
the mket
coverage
An ~mplOyee’s
upon
his
to compensation
in
covaage
horn the
offas
whi~
jOb s=r&.
23
a utihty
of other goods,
indirect
uttity
of utfity.
an tidivdud
is conditioned
(lg88),
.Compo.nent and a static
by &ooskg
-d
X,
fuction
O~ezs.
mtidzhg
given a budget
whi&
maps
Offers of aployment
values according
on his in&ect
he faces from MS cmrent
Z6S==~,. TOP,I (lgss), T.pel md Wmd (1992), MacDon~d
specific
cOverage as p.mt Of compensation.
is mde
by =
- =d
imm=ce
depends
rebtive
the
experience
desked level of hdth
coverage into a measme
job semch problm
of compensation
offers,
the job sem~
The resermtion
h employment
ad
level of consumption
consist of a wage rate and a price of health fiswace,
fmction.
(1979) md
of prices for health hsmace
health ins~ance
consumption
covmage,
wage rates =d
employer
u having both a dynafis
level of health insuance
The solution
of health hsmmce
offers on the cmrent job depend upon location
cost of providing
is modeled
of a job tige
fOr he~th. services.
in the form of health insuance
uket
Utifity =fization
dem-d
his cmr=t
Also, the offer distribution
ud
health insm~ce
by factors wbi&
actual =d
individual’s
the form of wage income.
in the employer
decision.
md incr=ses
an individual’s
of a given set of offers horn
and the evolution
se=&
on the one had
attainment
costs =d
for compensation
md
the price .of health inzwmce.
- education
specfic
employer
I adapt these mOdels tO include the
in my model with a reservation
ad
to ~
coverage.
( lg86).26
in the individual’s
decisions ti the model are gtided
- location
distributions
model of Jovanovic
(1978) and Morten$en
coverage
on the other by factors wMA
utifity
cment
of both an individud7s
for health insmmce
model is replaced
of compensation
variation
in~vidud’s
ad
Job mobfity
preference
offas from =
coverage
employ=
utifity value of both wage income
duation
both the value of health insumce
by shifts h the curent
sea~
coverage
for both a price of health
The model shows how the probability
value of health insmace
growth
sem&es
decisiom
as presated
tradition
individud
his job mobi~ty
The model is in the traditim
model
a
of health insmance
the model represents
is tiected
=d
md
model in wbi~
employer
to MS
utifity fmction
ad
the wket.
The fo~owing
employer
thee
sections
offer distribution,
ticmstmces
whi~
ad
wodd
outfine
the static
conswption
the search problem.
create job-lod
decision,
A comp=ative
and the ~mences
the dete~nation
of the
static analysis &macterizes
between
dyndc
~d
the
static models
of job &nges.
The Consumption
Decision
htividuals
a level of health insuance
&oose
To begh
suppose
that the utifity tiction
X
is a composite
increastig
comotity,
=
underlytig
for tiw~ce
relative
demand
u(x,
for medlcd
to other goods.
given a price s&edule
(3)
~d
a
E [0, 1] is a par~eter.
rate of substitution
care is represented
The budget
w=
constrtit
by employers
det=fied
consists
by a, which indicates
in the sear&
The solution
problem
of the problem
d’s
preference
(4)
stated in (3), (4) is specified
For tied
Thus,
response
in (p, w) space,
cmves
utifity
1 illustrates
a contkum
~,SPecfication
health &wmce
(w, p) offers dow
(~) i, ~attractive
is
as
utitity
the relationship
level,
between
Z, the fo~owing
resdts
in~vid-
apply
(6)
in w and p, with slopes that are increaskg
ad
bid-asked
cmves
steepen
in
insurance.
md wage location
the cmve
h that it impfies ti~tid”ds
AB.
=.
decision. for.~
intividud.
He faces
Foi a given “mmket price of health
risk-lovkg h regma
to ch=ges
coverage, p. HoweYer, none of the resdts below w~ depend upon this ss-ption.
24
~
> o“.
to the northwest,
for h~lth
the health insmance
of potential
accepts
(5)
(5) shows
are increastig
increases
to an increase in the preference
Figwe
indivi dud
we–ap
dzw
—
dpda
~>o,
dp
in a.
=
27 Equation
and wages.
For given utifity levels bid-asked
offer that a
Compensation
of the next section.
utifity ftiction.
for tismmce
the preference
is given by
The employment
u(w, p,a)
where u is the indirect
is
on a. The
XtpI
(w, p).
of a pair,
Utfity
between them depends
where w is the wage, the price of X is set equal to one, and p is the price of insmance.
offered
for insmmce.
I;a)
1 is tismance,
in both X and I, but the mmginal
indlvidud’s
other goods
is
u
where
ad
~ the Price Of
bsumce,
P, employers
locate themelves
but a zero shine of health
ad
ffi
hswmce
=d
p=tid
wve
coverage.
insmance
combinations
insm=ce
m
is tagent
coverage.
individud
Com~nsation
k
I motivate
distributions.
provided
the mgment
=e
an employer’s
of compensation,
&mmce
from =
There
of insmance
Although
the Est of mandated
mental health cme, physical
-d
tbe =s_ption
cost
pmtitiar
pa&age
(WO,PO).
cmrent
of a job mat ~.
therapy,
of offering
as F[u(w,P)]
mverage
health
insm=ce,
of utiEtifrom
or F(w, p),
Employers
below, yet their &Wacteristics
utifity off=s,
compensation
offer distribution
location
in inswmce
health inswace
for specific
-d
offer
(w, p) pti.
is given by F(y).
is a detertin=t
in the cost of employer
provided
treatment,
or F(u)
value of a p~tictiar
varies amOn.g states, ex~p!es
dcohohsm
fuction
The offer distribution
Employer
of state variation
inclu~ng
compensation
distribution
employer
of m individud’s
vmiation
-ket
(ti,p).
Ms possible
that W employer
treatments
ptis
to be the uti~ty
detertie
souces
coverage,
ad
provided
costs.
The fist
is
poficies must provide
kinds
of health
services.
Of SU~ treatments
midwife services.
since they =pmd
health
~clude
State redates
the rmge
of ‘usual’
price ,Ch=dulw fo, health tisurance coverage, because employer $i.e. the rate Of CmPlOY.C
chmacteristics of the state where the employer k located W tiect
that an tidividud’s
Mthough an employer’s stie md
&=acterktics
health insmmce
the compensation
may be represented
determinmts
are thee
levels
t~over,
offers =e
there is state-specific
mitim
z. ~mp,oyer~ face ~e,=nt
of wages
point C, where his bid-asked
sear& model presented
and the qu~ty
States m-date
the Mm
some combination
choose
is a probability
individud’s
eqtifibriu
state mandates.
ticrease
employers,
specific
lomtion
COverage.zs
off-
of the own mployer
Employment
of an individud
bemuse
offers the lowest wage, w(l),
of these wage ad
At C he receives
of F is intended
actors in the p=tid
h this model employer
-y
UO wodd
offer distribution
offers.
as wefl as &acteristics
among
cwve
cwve.
employer
on AB employers
a choice
the specification
the offer distribution
are p=sive
At A u
offers the highest wage, w(O),
Offer Dtittibution
i.e,, from dt=native
but k W mes
Sii=ly,
offs
A compensation
by employment
the m=ket,
Gjven
with bid-asked
The Employer
At B m employer
coverage.
At other points
to the -ployer
this section
on AB.
the employer pretim
observed state of residence k mog.nO~$ to fis h.~fi
tunovcr
rate me also hportmt
detertimts
fim=c=
employer. For that reason they are not comidered to he exogenous detez~mts
model.
25
cOver?ge chOice,.
of health hsurance
are correlated with other choice vwiables & the model, such as m kdividua~s
rate. I *C
price, those
dechion to work at a
of compensation h the
health
mre services
Stat=
reqtie
compaties
employers
compaties,
ewense
of .compfimce
insmance
Finfly,
ad
cash b~ances
coverage.
they place sm&mges
wit h a state’s regdations
pretius
WY
according
losses when a patient’s
incm
fimcial
loss is termed
and patients
‘compensated’.
Although
mu&
for Medictid
insmance.
md patients
are oftm
wable
Medictid
covered
by Medictid
charge higher prices for health services defivered
coverage.
This practice
health insurance
prefims
(to the extent dowable
to insmance
in the cost of wcompensated
Quahty
also vmies between
of c=e
services
is partially
as a tity with a relatively
I=ge
patients.
to provide
subscribers
tetiology
Z*~mp,oy=, ~ho ,e~.~,m,
compliance,
ad
Hence, state -dates
md
Just as with other industries,
by agglommation
effects.
share of the best smgeons
expensi%-e servicw,
the best trtied
=, “,u~II~ .,,mpt
bfls.
with
bsurmce
in the form of higher
reimbmsernmt
states, there is
employees.
fi.~ .t.t.mndates,
.o_ercid
Boston
in the comtry.
of c~e.
or catheterization
but se~-tismtig
ksmance
is w~
Mimeapofis
Hospitak
is
~m
h
for hemt attack
~e able to provide
Quafity of cme, howev~,
the employer cost of health tis”rmce
26
the scale and qutity
For example,
Mgh qutity
such ~ MN
not less expensive than puchmtig
may inme=e
1986).
differs between
States that have high quahty heduh care me ones whose hospitals
most advaced
thm
certtin
(see Phelps,
popdation
also we~ known as a city where patients receive a relatively
theti abifity
hospital
care.
states.
detertied
efigible
to individuals
by the mmket ). Since Medicaid
=e state specific and the size of the minsmed
state variation
(who ae
“to recover some of the losses from
is known as ‘cost-shifting’
pass on the extra costs &om cost-shiftkg
do
to tinsued
of thek
me
insmance
that resdts
usuaUy does not pay the
wcompensated
hospitak
@ort
of cme.
why patients
care is attributable
to pay the bduce
h =
c=e
reasom
expense for hospitals.
known
The
for employee
Hospital
care is a l=ge
of h~th
to
that are state specific.
Uncompensated
sdedules
compties.
the cost of qu~ty
there me m~y
of uncompensated
recipients,
mdy if they me poor)
care md
biU is not paid in fd.
who me covered by Medicaid
cost of treatment
compaties
Some states
to the cost of pretim
the cost of mcompensated
not pay the ftil cost of their treatment,
primte
States
paid by employers
to other costs of health cue
Hospitak
for Medicaid
of clh.
biUs ptid by insmdce
contributes
ones ~e
ftil
the prefims
tidmtry.
coverage.
employer
in fiancird
for payments
States t=
on hospital
The two most tiportut
patients
plan.zs
role with regard to the ksm=ce
to hold titim
from offering HMO
imm=m
health
in a health insuance
dso have a more tradit iond regtiatory
tiuance
prohibit
covered
the
is positively
may O~Y be lCSS~Pensive
h the absence of state mmdates.
coverage even Z m employer serf-b-es.
related
to cost,
wbtion
h
Hence,
state level diffwences
the cost of health
services
ud
k
the qu~ty
dttiately
of c=.e me a souce
in the preti~
of state level
price of health imwtice
potities.
k
fiport=t
=smption
geneous
across employers.
be=use
an kdvidud
in modek
Without
wodd
productivity
depends
su&
not =pect
haps the central contribution
of job
m
on the qtity
is that compensation
assmption
se=~
of Jovaovic
mobifity
there wotid
be no value to job se=&,
to gtin Mm a per~ent
(1979)
was to formUy
of the fit&
between
offers me hetero-
incre=e
introduce
in ewfiWs.
Pw-
the idea that employee
an kdividud’s
skins md m employer’s
tetiO10gy.30
Models
of job mobifity
why vmiation
k
compensation
Jovanovic’s
job matfi
ass~ption
-d
production
tetiology
heterogmeity
good.
have t~en
levels amoss
employers
on a ptioti
=d
labor
Tasks =e
production
bmdled
tito s~~m
efficient.
h
cefi~
sets of taks
according
is hired by a fim
wiU compensate
to compensation
to the efficimcy
with wbi&
worker becomes
h a job the l-s
may gains~
that hcrease
worker productivity
increase
less thm
assumptions.
of
reference
to
of the job
mat&
One, a fire’s
kbor
by doing.
With
to ledg
when completed,
cded
to perfom
jobs.
actual productivity
without my
produce
a pmticd=
of jobs md
oft ~ks that Aes
a partictim
the t~k
a fim
ca
productivity
h complethg
be completed.
productivity.
job.
productivity),
job.
Skce
his assigned set
However,
there is
at the start of a job tiom a &tribution
27
fitit
The more productive
a
Thm, over time, workers
. . . t&
the competitive
fim
ody
at the margin cmrent pay wi~
,.
empIoyer k deter~ed
becomes mre
and
more h m alternative job (with a Maent
ti their cmrent
(not potential
most
job based on his abifities
set of. tasks, because thae is a te&ologid
pay in an dtwnative
idea when m fi~tid”d
The nwbw
is more or less productive
~.~ & ~odel m ,~pIoyc,,s productivity at a parti-
of t=
that,
set of tash
theb potential
actu~~d
fip~cations
have mde
justification
retms
the question
has a very intnitive expluation
room there is for improved
rewwds
radody
ad
him with higher or lower pay; respectively.
my p=ticd=
thdr
whi& * &aw
two fild
to the orgdzation
for a p=ticti=
set of tasks) tha
potmtid
efist,
Howevw,
decreashg
set of tuks
To the extent that w employee
of tasks his employer
a natmd
two, there =e
is a I=ge
one includes is determined
kdividual
wotid
view towmd
as~etry.
that ea&
experience.
ody
the general idea of job -t&es
or ifiormation
an agnostic
Womds.
of offers reqtires
is Ned,
of ucerttity
A fire’s
Jovanovic’s
assmption
these two =smptions
cl-
aft=
of mt~es.
ti p=t by a -t&
Jovmovic’s
about tbe qutity of&
cOmPOnent
model tivestigatcd
cmremt match
OV.r
the
tke.
Here ae
two exmples
productivity
s&ool
in the cmrent job.
haeases
operator.
of kcreases
his potential
in his productivity
hcrease
h his potential
It fight
employer,
productivity
that have a lesser effect on actual
operator
as a fief,
over the next ye=
who takes tight
here is that in most
that occupies
it may be inefficient
the wtiter
the employer
tight
hthermore,
have to move to a different restaumt
Consistent
ako ticluded
with the fiteratme
as detertimnts
experience,
capital.
atttiment,
I represent
cost of health insur=ce
on etings,
-d
human capital
cost is denoted
there -y
&eady
given the fire’s
tetiology
of production,
h
the wtit=
in the restammt’s
his potential
h-m
capital
ex-pie,
efficiency.
ch~acteristics
Of ~
k my model
with the vmiable
by s. The distribution
over time
fis c~ent
jOb,
The wtitw
~~vidu~
a
status are used as kdicators
characteristics
be a
productivity.
offer distribution.
health
the -@d
him most productive?
a job wkich pays mOre th=
to re~ze
of the compensation
education.
the -gid
ad
job for the employee.
may muse a net reduction
as m
in a different job with the s-e
roses tasks me not divisable,
the SWS to be a mmager,
but to have two managers
cfiwy
manager.
assign him the set of tasks that mde
to create a dupficate
develop
for ten y-s,
in the job of wtit er may be less tha
is more productive
the dt emative job.
classes k
but does not increase his productivity
b the job of restawmt
that if m -ployee
then why doesn’t
employee
hman
First, a telephone
productivity
be objected
The constrtit
job
productivity
Second, for a wtit er who has worked in the same rest auat
timease
wotid
k potential
of utfity
K.
are
tidividud’s
of his level of
The state specific
offers from the uket
is
given by
F[u(w, p)] = F(%
Gener4
hman
K, S),
FK(U; K,s)
capital shifts the offer distribution
the offer distribution
to the left.
~(y)
=
where y is the next utifity
the offer distribution
haw
K,s),
~K(y;
of the curent
K,s)
from the cmrent
to the right, wfile
F,(u;
to the right, wtile
The distribution
~(~
<0,
<0,
K,s)
>0.
(7)
increases in state costs shift
employment
offer is given by
~.(y; ”K, s) >0,
employer.
Agah,
(8)
general hmn
capital
increases in state costs shift the offer distribution
shifts
to the
left.
Flgue
2“ Wustrat es a graphl:al
sation level =
employer
cost of ,compensation.
tit of health insmmce
derivation
of the .:mployer
locates in p, w space +ccorting
Both the average tied
V=Y across employers
utifity offer c~ve.
to the sh=e of insmace
cost of providing
health insm=ce
due to diffwences
28
in wployer
Given a :ompmthat @tizes
md
Ms
the price per
sizes md
employee
t mover
rates.
The figme
mce price s~edule
inswace
point
A, wh=e
*O
hw
coverage
coverage he must pay a Wed
he offers fti
hsuace
cost of offertig
prim
s&edde.
He locates
price ad
by the 10CW of pohts
The probability
coverage
Several imphcations
depen&
in health hsuance
shows offer cmves
offer cmve
is AzBz.
individud
health tiu=ce
price P(O). k
employee.
Mz, ad
is decreasing
The mket
h the employer
sh=e
draw from the offer distribution
on the demity
for two state.
k
of employer
insmace
model
ca
po~cies
insmmce
coverage
3a.
=e
higher h state 2 thm
h state 1-
the mket
the
individud
in state 2 is reflected
is below the offer &stribut ion in state 1 except for aployers
who
coverage to theti employees.
for health
insmance
1’s bid-asked
-d
The effect of
is i~ustrated in Figme
in state 2, PZ(0), is greater tha
3 b shows the effect .of the preference
of person
covwage.
at least a certain
be i~wtrated.
whi&
steepness
at
offers at diffwent points of AB.
h the offer distribution,
dem~d
covaage
tism~ce
wi~ contti
in state I, PI (0). The higher cost of health tismmce
a higher
faces a steeper health
of ffl
price of health insmace
Figme
Employer
state 1 the offer cmye is AIB1, mrd h state 2 the
Prices of health insmmce
do not offer health hsmace
w(l), p(l).,
utifity offer cmve is givm
costs on the ut fity offer distribution
price of hdth
order to offer my
He chooses to locate at
pa~ge
coverage,
one.
insnr-
he wordd pay a wage w(O), md
and a compensation
of the static utifity mtization
The figme
-ket
p
coverage
be the uket
a higher wage than wployer
AB, where
Employer one faces =
at point D, where he offers WZ, PZ, less insm=ce
that a compensation
of in.swmce
state v=iation
for two fires.
cost, MI, for ea&
coverage
tied
a higher tit
levd
to workers yotid
a high=
tiw=ce
decision
given by Cl. H he offered no tiwace
the price of fiurnce
hdth
shows the location
for health
cov=age
th=
hsuran&
person
on job &oice.
Person 1 h=
0, whidr is reflected
in the greater
CWV+ U1. He locates at point
a lower wage tha
do=
D, where he receiv=
person O, who locates
more health
at point C (P1 < @o and
WI < We).
Figme
the fi~e
3 c shows the effect of the qu~ty
m indvidti
the mmket
as a whole.
offer distribution
~pect
ti&vidud
p,
w.
h
=t
a low value of the employer
He is lo=t ed at poht
& on the offer &stributions.
h
matting
to
C, where he receives
in compensation
by swctig
3 d shows how pre-etisting
health
with a pre-etisting
condition
the absence
health
of a pre-tistiig
conditim
29
mat&
rdative
the ~ket
component.
He cm
for an alternative job.
conditions
h=
component
w, p. Memwhile,
is given by A lB1 for mr average value of the employer
to gtin an increme
Flgue
is
has &am
of m employer
-y
located
resdt
at pobt
in job-lo~.
k
the fi~e
an
C, where his mmpmsation
his other job opportnuities
wmdd be given by
the cmve
Al B1.
health insuace
wodd
However,
the cost of tceatment
,poficy, so his actual mmket opporttities
prefer my dt ernative
his pre-existtig
of the points
for his condition
condition
compensation
offer dow
has lowered his mmket
along the cuve
AzB*.
k
the cmve
this sense kis cmrent
30
be excluded
from a new
me @ven by the lower cmve,
opporttities
cmrent job.
wodd
AzBz.
Al B1 to the left of p otit
He
C, but
su& that he prefers point C to my
state of health
lo&s
k
tito
his
The Search Decision
This section provides =
of job mobfity
(1986)
=d
=d
outfine of the implications
health insumnce
Mortensen
(1986).
The model
utifity here dep en& on the due
wage kcome
done.
the traditional
hsuace
demad,
apEcit
model.
health insm=ce
difference
level of compemation
modds.
exmples
is sM=
s-ch
cmrent
~
compared
because
coverage
it &ges
(1986) =d
to the fist.
aployer.
Jov=ovic
kto
tinstrm
by Topd
~rst,
as opposed
to
none of the predictions
of
treatments
of &wes
job-lo&
presented
here dews
compmsation,
k health
effect cm be ~de
Dyntic
mat~-specfic
~on~tat
yomg
work=s
productivity
-ter
youth~.3Z
across fires
Upon
and severd.other
behavior
modeled
in those
The model of Topel (1986)
or explored
con&tion
cwrent
employer’s
in his mod~
&ages
remtim
The
in the level of
relatively
thm
for
=
contained
location.
,e=&
is a de fato
tractable
one ae
good enough.
in =
of v=iation
expected
which
the distribution
in
is a fied
fOr a new jOb ~ the
Job offers from the
on individud
health hsmace
horn
in ititid
chmacteristics
offer from the mmket is tidependent
10cation though
beha.ior song
&aw
Heterogeneity
of leiswe,
WM se=~
so that conditiod
However,
me.s
the utfity
somce
individual
the cmrent
across locations,
cmrent employer’s
not ~m
a job
receive a compemation
capit 4 &=acteristics.
and individuals
~c=ption is the study of Lancmt.r ~d
do=,
md
their hman
accepting
a better job
distributed
m=ket,
as long as it exceeds
levd of health insm=ce
upon an indvidud’s
~odel
ordw
~ the
not new to on the job
effect of apected
for the fist
the kbor
upon
The offer is accepted
market are tifordy
kdividud’s
dyntic
fiect
molds,
The dynatic
Wects me not frustrated
wtile the expression
for obttiting
the expected
effects =e
(1979) describe dpc
for a sfiple
of offers conditiod
for ~
prospects
for. a dw~c
relevat.
the model
wage offers.
I ~ow
appfied work, howev=.31
to the model here, but dynadc
the distribution
~h=
ofly,
supply, md wage% the theoretid
offered by a cuent
not been adopted
empiricfly
s,
deterti~tion
from models
health timmce
it does Wow for sepmate
is substantial
Mortensen
model
,, ~
model is a&pted
of SU* models have. app emed fi the Etwatme.
papers h=
ad
for the jotit
here differs from those models in two respects.
is cosmetic
However,
betilor
within this fimework.
The second
sear~
The se=~
of both wage income ad
The Wwence
se=ch
coverage.
of job sem~
coverage
fmction
of m
win depend
of compensation
(lg84).
h&vid.&
the.
31
who are entefig
the labor m=k=t fOr the fist
offws from m individud’s
The setup
ad
cwrent
solution
employer.
of the search problem
the model
is continuous
individud
at my time t is given by O(t).
uket
and contenting
by the cmrent
md
the job se-~,
-ployer.
The fist
where rt* is the res=vation
period
is represented
value of s-ch
Ftially,
ordw
fuction
utfity
of utifity
employer’s
~(Y)dY.
n=t
d E[Vs(y)]/
h the next inst at,
Equation
h
(10), both anticipated
reswation
dE[Vs(y)]/
is strictly
offer from
the market;
ad
This assmption
~ compensation
I asstie
in y, i.e., the value of staying
fipfies
with respect
r is the discomt
F(u)
= f(u)du,
fmction
34Tti~
.,~”mPtiOD
c is the perto time of the
aployer,
rate;
u is the
of the cment
md the =pression
d~(y)
in.the
compensation
that an individud’s
from the cmrent
uti~ty
employer
offer are reflected
that the value of continued
at one’s curent
that .wiU be offered
reservation
job
md
wage is ticreashg
semch,
contenting
by one’s
fiect
the
fi the ‘em
c~ent
in expected
VS(Y),
to. semd
emp10yer.34
shocks to
utfitY cOst Of ‘C=*
futme value of search. It is ako a clarification of the model of Topel (1986).
i,
,Cquued,
b..au,,
ti.reas.s
h
cmrent compensation reduce the e~ected
That is, bcreases in cmrent compensation reduce the probabfity
net ga~ ~Om scmch.
that m offer from the mmkct WU exceed the cmrent
offer. The assumption tip~es that the negative .S.. t of .mrent compensation hcreases on the expected net gti
semch me outweighed by the positi”e effect of those ticreases on the utfity
stays h h
=
(7’) in Topel (1986).
$,T& i, ~ d,partm, &om Mo,t,ns,n (lg86, section 2,1), who holds constmt the net ~rent
and the e~ected
offered
is the is the distribution
(1986) and equation
~ho&~ to the cmrent
dt on the right hand side of (10).
for a new job is increasing
dF(u)
package
compensatio~
is the distribution
(2.8) in Mortensen
actual &anges
utitity level. 33 ExPected
increastig
a job offer from the
offer horn one’s cmrent
the mmket;
~(y)
to m
(5) into (g) gives:
to equation
wd
job
the mmket;
offer, y. The expression
equation
(10) is stim
given a new utifity
b
is given by
dt is the derivative
offers from
of a compensation
compensation
Substituting
for the problem
the
atiable
of rejecting
utifity level; Z is the utitity of cmrent
offers horn
Briefly,
The set of Mor-tion
Vs (g) is the v~ue
condition
y; A is the arri%-al rate of employment
instantaneous
by t E [0, m].
in the appendix.
given that y wi~ be the new compensation
cost of search; the expression
expected
are given
cmrent job.
32
from
that wotid be recei”ed K m titividud
compensation
on the
cccrrmt
the shock is positivdy
rehted
positive
on a cment
general,
job,
a positive
reservation
( d E[VS(y)]/
dt >0
utifity level, u“(t).
if not rtiected
utfity
in the distribution
), the strength
Thus, rm uexpected
price’of
of utifity
of
hdth
tisumce
offers from the mmket
in
level to inmease.
Statim
Expressions
for compmative
Station=ity
is kposed
stationmity
static effects me eufiy
on (10) by asshg
is imposed,
equation
the appendix
derived if stationmity
that the cment
is imposed
job qtfity
on the model.ss
offer is a Red
constmt.
(9) reduces tm
; ~:
—c+ —
U* = tie-up
h
shock
wage levd or negative shock to the qected
muse the reservation
Compmtive
For
to the cmrent
sho& to the =pected
coverage
mm
job.
I use (11) to derive the foUowkg
(u - U*) dF(u)
compmative
(11)
static ratits:
(12)
These
results are what we wotid
of sem~,
relative
Thus,
the cwrent
to othw
goods.
The insuuce
of resermtion
r, because
decreashg
in r.
Pre-ezisting
A special ue
job moblfity.
utifity.
lasms
The reservation
utifity
ad
the import=ce
level is deer-tig
the preference
inswace
to the price of insm=ce.
utifity level is dso
level is increasing
in the cost
for health
of health insuruce
a prospective
A. E offers mrive frequently
utitity level is exceeded
utfity
effect is proportional
The resermtion
preferences
decreashg
job relative
the kter=t
to conttiting
in the cmrat
(Mgh A) the expected
b
employw
lmgth
M a
sewcb
wage ad
of time mtfl
is
the
=Y
is short.
Conditions
of the station=y
h
indvidual
model i~ustrates
with pre-etisting
loss equal to the cost of tre+tment
,s ~tho”gh
prderence
the present value of acwpting
offer” rate of mriml,
given rcrervation
The reservation
price of health tismmce,
a low price of health insu.mce
deterdnmt
rate,
employer
expect.
the ifiumce
medical
for his conditions
conditions
tigns.
33
c=
when he moves
~=rivation of .Omp=etiye static @ects fi.m (10) &ectly
none of the derivatives wodd have Ue.ent
of pre-etisting
ewect
medicd
conditions
to adme
a lmp
to a new employer,
wotid yield a ~.r.nt
on
s-
because
set Of exPressiO~.
pre-tisttig
conditions
we =cluded
the treatment
cost of ~
The exclmion
of pre-tisting
an additioml
cost to &mging
For Uustration
depends
individud’s
has a negative
suppose
employment
distribution
employer
Thus,
conditiom.
of a is tbt
h
k
utifity
A(I)
of utfities
offer distribution
offered
does not.
by the maket
Specficdy,
the
(13)
&aws
= m.
Positive
values of p shift the maket
offw
with a high dad
for health insmmce
coverage
of u than an individud
with a low demand.
Here my
it is representative
of the treatment
(13), a high value of p is associated
(13) impfies that equation
because it introduces
= F(w + p(a)),
an individual
is less Ekely to receive high market
intmpretation
effect on job mobitity,
loss.
offers is given by
> 0, P(O) = O, md
to the left.
The Weatm
the greater wiH be his expected
that the &stribution
G(u)
where p z O, alp/da
conditions
po~cies.
jobs.
of this pokt
of uket
in new health ksuace
pre-etisting
conditions
on a, but that the cmrmt
distribution
from covaage
cost for m individud’s
with a low expected
(34) from the appendix
C-
be rewritten
pre-etisttig
value of job offers. Equation
as
thiscase
–Fti
8U*
aa
Comparing
from job search.
level we c-et
because
for health insmmce
Mthough
itier
the probability
The probabi~ty
– A#=(l
T+ A(l–
=
(15) to (35) in the appen&x
Now a high de-rid
e-a~r
a high due
– F(u*))
we can see that the negative
coverage
is associated
of a is associated
effect of a on U* is re~orced.
with a low expected
with a relatively
utifity gtin
low reservation
with a high probability
a good job offer from the -ket
of an efit to a new employer
(15)
F(U*))
that a high value of a is associated
of receiving
< ~
utitity
of job &age,
is low when a is tigh.
is
.(16)
Hence,
the probability
of m etit decreases in response
to m ficrease
ar
z
‘-’w(%+~)
<o-
34
in a if
(17)
Since 8p/~a
Equation
coverage
(17)
the relationship
offer distribution
de~d
his cmrent
offem
condtiom
w~ng
his Aue
enough
job
th=
to accept
coverage my
reduce
insuance
this ass~ption.
of a is high enough,
due
utifity
if
to the de~d
has a greatm
level.
for health ksu=ce
effect on the memr of the
E so, a
kdividud
with a high
seek a job that offers more health insm=ce
coverage
offered
that job
coverage
the probabfity
related
coverage
on the res=vation
(even W the new job
of health
wodd
is negatively
for health ~smace
for health ksm=ce
if m=ket
h eq. (17) holds if md ody
shows that job mobfity
when the demd
uket
th
>0
are scmce
of a job
An individud
stice his expected
a low=
efit
wage),
for him.
net utihty
35
The aclusion
for some kdividwls
wiH be discomaged
of a.
but he may not fid
gti
of pr~etisting
even if we were not
from job semch altogether
from semch is negative
if
for a Mgh
An
Empirical
Analysis
h tfis section I retmn
question
of the
coverage
deterrent
Search
to the empirical
posed in etisting
a greater
Job
Model
issu=
r~sed
studies of job-lock
tO jOb mObfity
is whether
fOr ~
in~viduaI
thm for one who does not. Although
of whether
or not the empirid
mary pohcy
mob~lty
implication
in a regtie
of employer
conditions
health immance
~ed
coverage
coverage
evidence
intervention
in the market for health insmance
o~y a tinor
improvement
to the acquisition
h~th
insur=ce
regtie
fided
health insmance
of employer
that exist ing studies of job-lod
not been empiric~y
The empirical
pfimtions
identified
model
as weU as the expected
probability
of a job
that probabi~ty.
An objective
conditions
me a deterrent
then govwment
compaties
to job
intervention
from acluding
pre-etisthg
to cent inue to offer the compay’s
suggests that this stop-gap
cmerage
wodd
approad
in the most opttistic
coverage.
Ody
a broader
pubfic
can address the inefficiencies
It also .qu~tions
cause
new obstacles
intervention
in
caused by the cmrent
the estimation
sugg=ttig
to pubhc
scenmio
and would more Ukely introduce
coverage.
below
hig~ghts
that the job-lock
that expected
Previous
ch=ges
nests the basic model of previous
=dys=
c=
of tisting
in a cmrent
emphasized
~~ysis
strategies
effect has
mobfity
the empkicd
im-
studies of job:lock.
employer’s
compensation
offer
me key dete-ats
of the
costs as key deterfinmts
wi~ be tO specify ~
as a special
job-lo&
between
used in previous
offers from the m=ket
modek
of the empirics
be used to test the adequacy
the sharp contrasts
and those of of the models
value of compensation
&ange.
ins~~ce
in past work.
model
m~del. imphes
he~th
are
n?ne take issue with the pri-
coverage,
have used in their malyses,
specified
of the job sem&
The job sem&
wtich
coverage
conditions
is warrated.
health insurance
the form of national
medical
ksmnce
to the rate of job mob~lty,
of affordable
medical
studies differ in their =sessment
this hypothesis,
or forces employers
to former employees
I present a new set of empirical
of efisting
insmace
which either prohibits
from treatment
or not pre-etistkg
I pre-existbg
health
RecaU that the empirical
.whO has emplOyer fi~ed
authors
supports
of the hypothesis:
in the form of legklation
medicd
evidence
in the introduction.
empific~
case, so that .esthates
of
mOdel wfi~
of model parmeters
models.
Data
The data used in this adysis
smvey of 12,686 yomg
21 as of J=u=y
is from the Nation~
men and women h the U.S.
1, 1979, in the fist
Longitu&n~
Rwp ondents
yem of the survey.
36
Sm~er
of YOuth (NLSY),
were between
The 1979 NLSY
s-pie
a p~el
the ages of 14 ad
is representative
of the U.S. population
-A
since 1979. 38 ~terview~
ye=
smple
in the given age group
extracted
have been attempted
for this adysis
perths
The NLSY is the ody avtilable
dynamic
interactions
onfy available
&ages
and a sid=
includes aployer
of dat-
identifiers ford
aployed
respondent’s
in mch job,
job rbmges
=d
At ea& fiterview
insmance
&=ges
construct
a discrete
(rougMy
offer question in conjmct
point-in-ttie
status,
location
offered
predictions
of a respondent’s
and codes fOr m in~vidud’s
ifio=ation
is partictiarly
usefd
I use the codes for an individud’s
s.At ~he tie
37s_p~e
is bow
coverage
emploment
as 1978.3s
coverage
if not employed.3e
or not a ptiictim
using the aswers
to the health
identifiers,
record of individud
coverage
eqerience,
state Of residence
for generatkg
record of a
status =d
edncationd
of health
to -t&
-A
to
wage rate.
atttiment,
at each i.nte!view.
predlctio?s
state of residence
characteristics
md
The residenti~
tismace
h&vidud
coverage
with state
&tory
for one or more previous titertiew yews. Tfi
data sets. Udm
pr=tice
accowts
respondent s-pie
retied
for the low attrition
h the SUVCY.
can be recovered.
,U,ey
ye= except for 1981. For that Ye=’s fiO~atiOn
status when possible ustig other Nomation,
of the pr.viou
longitudtid
to be deceased, titerviews are attempted with each respondent each smvey ye= even X a
ssTh=t ~“=,tion h= been ~ked k .~
the i~
ba&
of whether
health insmance
rate h the NLSY. As of the 1992 SWVCYyear 91 percent of the titid
etike emplo~ent
cent huous
record
.f ~ti ~rit~g, the NLSY,, Iggs suvey year ktertiew romd had been r=centlY fielded.
respondent was not titerviewd
k-mce
record
cfig
or not he is offered health
attrition ~ the NLSY h= been ,hgh$ ~elativ= to .ttrition ~ other popdm
a r~ondent
a -t
or most recent employer,
enough longitudinal
of m individud’s
ad
The NLSY
health hsu=ce
was asked whether
ion with the employer
a rifi
of the s~vey
it is the
employmat
covaage.
m approximately
longitudinal
analysis of
beuuse
respondent’s
record of an individud’s
=ual)
The ata
y-s.
job &mges,
status at least as f=
if employed,
atistered
needed for a
record of =A
him health inswance
data also conttins
The data includes measmes
status.
in employment
since 1979 ea& respondent
insutice
The NLSY
the itior-tion
coverage =d
construct
been
for each respondent.3T
changes in health insur=ce
a longitu&lnA
of a respondent
health
respondent’s
so that one c=
employer
obttin
in health insmance
coverage by his cmrent employer,
One -
conttins
ha
1992 int-view
jobs held stice the begiting
It is 4s0 possible to constmct
states.
amufly
cent tins both a longitudinal
record of eafi
The smvey
to the 1979 though
data set whi&
between &nges
data set whifi
at that time.
I =scert&
such as whether a respondent rectived health bum..
htervicw md whether that previous employer is the c-rat
37
employer.
he~th
at
level measmw
individual
with his statels
coverage”
measme
of the cost of health insmace
ad
“average
“percentage
hospital
room chage
coverage
(whi&
of the NLSY
in a study
of job mobihty
ad
demd
NLSY
level of health insuace.
-ong
27 to 35 years
poptiation.
Within
this young
is probably,
if not cert~y,
because the ticidence
of health insm=ce
old,
ovetight
I ~t&
health
each
insm=ce
stay ’’.qo The fo=er
is a
the cost of health cme
is f=
of the NLSY sample)
health insmace
respondents
&
of ihess
coverage
-d
is Ekely to be below
of their 1992 interview
does have a
of NLSY
ever having
received
tim=ce
m~ogously
defied
job mobifity
in a population
proportion
coverage
is a dec~ing
addts
fuction
of my sample
insm~ce
the
U.S.
coverage
cross-section,
expenses.
utifity
Hence, if the
in the U.S. labor -ket,
jobs
several
tties
without
is Ukely to be higher than the
in the U.S. labor force,
because
the rate of
of its average age.42 Health insu=ce
ushg the wud
M~.h
coverage
—.
~terview Of the C~rmt
room chmges were repo,ted h The Source Book oj Health
Smvey (CPS) from 1979 to””1992. The hospit~
Association
respondents
the size of that effect.41
who have chmged
~0The state ~~th in,ur=nce CO,,,ag, p,,mntages were cal dated
Insumnce Datq Health burmce
medicd
from an employer
of non-elderly
NLSY
the national
effect on job mobfity
respondents
the average for
of age md the mmginal
of ~pected
sample is fikely to underestimate
the proportion
health
ftiction
function
The average
of the non-elderly
the average ut ifity value of h~th
is m increastig
coverage
yea
below the average for a cross-section
disease is m increashg
apply to the use
coverage.
so that even in the last smvey
age group
analysis based on the NLSY
Popdation
with private
much lower than its average value wittin
for health insmance
Second,
&anges.@
U.S. popdation.
average age of”NLSY respondents
a
particdm,
the latter me~mes
relate to the youthfdness
of the non-elderly
in age from
dem=d
ad
h
state.
two mveats
raged
population
for a siWle tight’s
However,
a cross-section
i.n his state.
of non-elderly
of the su>ply of health tisu~ce
h m individual’s
coverage
of Ame~ica, 1979-1993..
—
4, A,
~
~“~cetion
to
~fi~
qufica~ion,
~her=
factor h job mobihty deci<ons eve” with
-ried
de.
=d
either have .Wtien
ts the bportance
both wedti
-d
hcome,
so yowg
thm me thek older comte~
titfi
a popdation
tiportan.e
42 Job,
-ong
~tich
.fler
NLSY sample. A &&cmt
to have cM&en;
thek f~es’
coverage ~ a signficmt
proportion of the respondents =.
ewected
medicd e~enses
positively
coverage h their choice of employer. Second, age is positively related to
individ”ak
arts. T&d,
to believe that health tismmce
r.m.m
this youg
or =e plmning
of health tisuance
is
me less able to &aciaUy
the ticidence
tithtmd
of actidents and tijties
cross-section, so the classic tis”rance
mexpected
k higher _ong
role played by health timmce
medicd
e~enses
the yomg
thm
coverage may be of equal
the yomg compared to a cro~s-sect ion of tidividuds.
h=~~h
~,wmce
..v,,ag.
m.
1,SS
~re.d.nt
38
witti
the group of jobs .cmpi.d
by Yo~g
wOrkcr.
is dkely
to be m fiportmt
f=tor
never received
health hsm=ce
the probablhty
of obttig
his past inabifity
on the NLSY
mobfity
in the decision
covwage
a job wMA
to obtain employer
win uderstate
from
Efied
Haled
tiects
a Mgh@
health
hsmmce
for w
tidividud
he is Wely
coverage
coverage.
between
benefits
As su&,
thm
do=
to b&eve
that
analysis based
covwage
of r=pondents
category
who b
is low based on
a
health insumce
pwcentage
hsmance
jobs
because
health inswmce
the interactive
of employer
employer,
offers health
to the extent that the NLSY h=
low probabihty
m
to h=ge
ad
job
h the highmobifi~,
the U.S. labor
force
k
general.
Sample Sel=tion
I use ody
the cross-sectional
subsmples.
suvey.
and Descriptive
Statistics
saple
of the smvey,
There were 6,111 respondents
Table 3 shows how the respondents
select ion criteria
accout
s-pie.
selects respondents
The fist
age of 21 years old.
=
of job
analysis
to mde
decisiom
inswmce
coverage
for &ost
mobi~ty
who my
of this selection
decisions.
independently
be completely
respondents
who have been interviewed
Wormation
is avtiable
excluded
criterion
was ‘keeping
house’
selects respondents
no more
thm
for a more severe rtie - select ody
from mdyses
of job mobifity,
the dud roles of ‘wage emner’
thm witti
for the propose
requirement
point
of their health
of edorcment
of respondmts
despite thek age, ad
agents.
fox
must be old enough
k te-
=bitr=y
the
for
at the lower age
the k.elusion
of
The selection rde also chooses
This is intended to ins~e
of contro~ng
that enough
for mobserved
individud
h the mdysis.
interview
substitute
tidependent
Two
the cross-sectional
A ti-
especidy
the =clusion
eight or more times.
for ea& respondmt
The second sdection
m
to be considered
yem of the
were selected.
made from
in the analysis
-
of their pments
in the titid
at least eight times titer rea&g
included
between
mrd supplmentd
in the adysis
of theti p=ents,
independent
who =e too yomg
saple
was twofold.
The age of 21 is atittedy
to strike a bdwce
the tifit=y
of the =clusiom
who were interviewed
respondents
chm=taistics
who me hcluded
is that the tidividuds
sigtificmtly
both
the cross-sectional
90 percent
The titention
this role, but was &osm
margin,
b
exclutig
workers bcrewes
one time.
mde
beaer/home
tire’,
activity
This tie
respondents.
as thek mployment
md ‘Wd
the group of jobs occupied by ~
whose mtin
was chosen
Fem4e
patterns
in the week precebg
to be a filder
respondents
~e
usu~y
me thought to determined
where=
by
the role of men is thought
workers b the U.S. labor force, because the Mgh job mob%ty of yowg
the employer cost of offerhg hedtb &mmce
39
coverage for the jobs they often ocmpy.
to be more cle~ly
in the NLSY
labor
force
attempt
shgdar.
sample,
rate
used
interviews.
of labor
the cost
lnhosp, -d
md
age.
in i’s state
pctchghp),
of time
4 shows
consecutive
much
status
transition.
employed
either
in jobs
the next
interview
have
or he codd
qtit
*aJob ~obfiity
fit ertiew day, k
i’s hma
thee
percent
coverage
a job-lock
for job
stayed
he codd
have
in his titid
of job
left his job
period
job.
Fow
status,
2, because
its mean
at least a \veek out
Atittedy
tidivid”al
44Altonji ~d W=.-
(1993) found the proportion
Dyriafics
of 36Yofor titividuds
to be approfiat
of the
effect.45
inswace
transitions
type of job trasition
(period
cotid
have
of health
md
not h the titary
health
t) respondents
do not.
ocwed
due to a layoff,
between
- a respondent
&stissd,
insurance
are
By the time of
or fiin~
coverage
is coded as havtig left a j ob if (1) his employer at the next titertiew is Mezent
is coded as stayhg k his curentjob if he is employed at the s-e
is a
this low percentage
is .Onsidered for respondents who are employed, not h school, and not ti tbc fit~Y
cmrent employer, or (2) the respondent is not employed -d
Study of hcome
types
fOr SeP-in~
( illspell and child), marital
or in ones that
involmtar~y
‘f
pcf-ch,
in i’s state
in table
a possible
tr~sititins
28%
(hi-pet,
and heath
covaage
1, =Cept
ad
gmde),
of the two interviews
insuance
types
out
mems
md
insmmce
of the cases.
percentages
t + 1) ttiee
is m
rougMY
=pe~sgr,
( ezper~en,
from work due to ihess
for eittiatkg
health
and a high
their
~easwe
in table
capital
stands
2 reports
43 By t~~
defied
of health
At the time of the fist
(period
his job;
have
h
smple
that offer
1, and table
mobfity.
Each ceU of the table represmts
interviews.
insmance
codd
job
ills~ll
was away
variation
the weighted
in table
the cost
the v~iable
fidivi dud
fafi~es
role is to be a wage earner from respondents
for health insmmce
interviews
very
are defied
( unemrate),
e~ner
mwket.
of tidividual
vaiables
between
does not provide
labor
wage
apphed to respondents
of home makers horn the adysis
44 The rest of the v~iables
i’s demand
low 3.2Y0. ” h
Table
whuse primary
leave indicates
meas~es
Of au these
relatively
period
represent
up in an ma of dual
The ~clusion
for women.
in the analysis
the jobs. end between
sep-qwit,
W grown
The v~iable
deviations.
should not be so broadly
this concern
role is in the non-wage
The variables
~d
have
out respondents
to separate
stmdad
as they
participation
whose prim~y
However,
at the next kterview.
status
a. Of the
thm his
An tidivid”d
employer at the next hterview date.
of ,epar&t ions for mY reason b thefi samPl, from the p~cl
ely 24% (Table 2, all experience levels), bnt fomd the figher n-be.
with between 5 and 10 yeas of work expedience. h my s-pie
the average experience level k
7.45 ye+.,.
4ST&
~ercent.ge
is
status on job mob%ty.
10W
OdY
k
the
sense
it
proyides relatively httle vmiation for meastig
the effect Of he~th
for this age group.
It is not low b comparison with outside national statistics
40
tr~itions
we
possible
- had
hdth
insw=ce
fisumce
at t, did not have health
tiswauce
at t+ 1; did not have health ksmmce
there ~e
twelve possible job md
Table
h both period
-rent
of ~
job
t md period
(job stayer),
by the the
kvolved
job Ieava.
lost between
ones, though
off, tied,
it is possible
not perfectly.
t + 1. It is used as a measme
expect
b period
that amo~
to tintain
fied,
hyed
theti h=lth
off, or distissed,
not a new job wodd
fidividud
period t -d
m importmt
health inswance
in ofly
*TTh= ~~t~ction
b
hmlth
50.9%
insurmce
an individud
job
k 8.55%
job stayers ad
whi&
wfifi
involmt=y
w=
went from
in period t +.1 occmed
job leavers.
separations
individud
and volutuy
reported
tt 1. It is used as a measme
qtit his cment
who between
period
those that volutmfly
behg
layed
of involmt~y
job between
t md
factor in the decision
69.11% of the =es
where a
thm
to &age
period
t + 1 &ted
were able to -int&n’
individti
involwt=fly
be more
those who were
jobs is whether or
Table 4 reveals that of W cas-
coverage 70.970 finttied
hdividuds
period
left their job wodd
status h the next period
where=
in which m
thek hdth
immmce
their health tisumce
left a job that offered
coverage.
,6T~W there i, .=a$on to question the hphcit
coverage at a -rent
who stayed in his
job mobMty.47
coverage.
qtit a job that offered health insmace
coverage st atw
~
equals one if =
coverage
offer health insm=ce
coverage status in the next period,
separation _
between
insw~ce
h=lth
job stayers and 59.8% tivolved
those ~dividuds
sfice
in the table.
but no longer had health
t to having health hsmmce
coverage
at t t 1. Th~,
offered
interview
health
at t, had h~lth
a respondmt
ha~ of the cas~
of volutary
insmace
whifi
49.1% tivolved
equals one if an individud
from a job that offered health inswance
Mely
at the ititial
sep.ino
from his job between
insm=ce
who left MS job (job leaver).
Of those a-
to distin@sh
The variable sep.quit
We tight
a respondent
best
.The variable
at t + 1; had
represented
had a job
&ange. 46 ~an~ition~
no job
coverage
tr~sitions
81.7% tivolved
Of those cases 40.2~o tivolved
or distissed
job mobifity.
t =d
invol~ed
no health insuance
status
had health inswace
tht
insw~ce
did not have health
a respondent
and 18.3% involved
It is kterestkg
titerviews
the NLSY
health insu=ce
of the c=es
health
at t, did not have health insw=ce
of the second kterview.
k 8.09% of the cmes.
b
at t +1;
t + 1. Of those c=es
the cases a rwpondent
tisumce
hatig
that in 72.5%
4 reports
insmmce
at t, had
~s”mption
~ etisthg
studies of job-lock, that heal*h rnSUmCe
job k fied.
job reparation k not de--at”
between a job qtit md m tivolwt~y
coded m a qtit if an tidividud
left & job -d
layoff, diamis.d, or fired.
41
k the NLSY data. A job
cited u a remon for the s~aration
mytfig
except
For the s-e
reas.om
efited
from
a job
that
obt tin
a job
that
offers
dso
appears
to be
offer he~th
coverage
period
ad
true
coverage
data.
Among
of tidividuals
coverage
coverage
59.42%
individuals
tisu~ce
insuance
in the
I esttiate
than
those
individuals
obttined
a job
who left their
a model
were =ttited,
who between
job
who
period
qtitters
who involmt
in period
in period
of the tramitiom
one for those individuals
t and another
the outcomes
that amo~
consecutive
wotid
be more
hkely
arfiy left theti job.
t quit
a job
t t 1 that offered
t job
periods
that
to
This
&d
not
health
insumce
Iogit.
Two logit
involutmfly.
Look At The Data
For frustration
equatiom
health
53.17%
A PTeliminaq
expect
did not offer h~lth
insuance
vffsus
we tight
for individuals
h the prior equation
the outcom=
in table. 4 by mtittiotid
who did not have health inswance
who did have health
are those trmsitiom
in the latter equation
inswance
that appear
are those transitions
coverage
coverage
h period
in the fist
col~
t.
in
Thus,
of table 4
that appear in the second col~
of
table 4.
The resdts
of the fist
mdtbotial
Iogit (no health insmace
in table 8a. Tables 5a – 5d..show predictions
trwsition
health
horn a job that does not offer health insmance
insurance
probability,
=e:
of the probability
coverage
condition
job qtit,
f + 1. Each COIW
on the type of job trmsition
involuntmy
level of a measme
in period
job sepmation,
that an tid.vidud
coverage
in period
of ea&
75th quartiles,
sample.
The fist
CO1-
4. ~ do
~~,
~1
.g~i+g
of the table shows that the probability
,,ery
~...;i
h
the
coverage in period
,-Pl.
Possible job trmsition
of the state popdation
the dti-,
.of state private health insmance
job that does not offer health imm~ce
that tr=sition
types
the
costs.48
The rows of the table represent
and the maximm
a
t to a job that o~ers
k the rows of these tables I vay
h table 5a I VZY the value of the variable hi-pet, the percentage
bY private health insur~ce.4g
wiU experience
table represents
that occmred,
and no job change.
of state health tismance
coverage in period t) me reported
the g-e
the 25th, 50th, ad
coverage
of Wig
covered
rates horn the CPS
the tramition
t to one that do=
in period
from a
t t 1,.given
value for a particdar cost measme. I we the estfiat.d
model pmamet ers to predict the transition probabfi ty given the ~signed level of the cost measme.
4, private
h..lth
&sunac.
coverage k a respondents
the estkated
cdctiated
COV...g,
rates
may
of the employer supply cost of h.dth
~s~ance
state. States with lower snpply costs me Wely to have tigher coverage rates. Here I we
p~oportion of tbe stat. population
ustig the Mach
be a measue
that is coverd
extract of the Cwrent Popdation
42
by private health tis”rmce.
The percentages were
S“rYey (CPS) for the ye~s 1980-1992.
that =
hdividud
qtits
his job,
is hcreaskg
by health
hsnz=ce.
The second
traition
probability
in this COIU
Iowm thn
for job qtits.
does not appem
h
colm
cofim
ticr=es
the third
that a
For A
types of job tr-itions
in~vidud
age percentage
chmge.
males
h
level of hospital
tidividud
and mud
6d report
6a – 6d con~tion
The resdts
the tr=sition
related
ktwviews.
to the cover-
when thwe are positive
when there are negative sho&s.
=e negativdy
prices.
on the prob-
51 Again
job sepmations,
cost m=sue.
related
to both
we see tbt
ad
H=e
a state’s
the trasitiOn
that the effect of state
for the forma,
the predictions
for the probablfity
t that does offer health insw=ce
tism=ce
coverage.
the probabi~ty
related
sol lea,= the ~xcep~ion.Cp,,.,nt.d
of sdt.
the supply effect
that an individud
mvwage
As in the previous
wi~ expmimce
to a job in pwiod
tables,
the COI-S
that OCCUS, and the rows condition
h
t+ 1
tables
on Ievds of the
cost measwes.
that does not is positively
_,
than for involut=y
of tables 6a – 6d =e
of exceptions
not sigfiut,
status trasition
on the type of job traition
state health ins~ance
grti
the
but is clemly
status between
is positively
coverage tha
change in hospital
from a job in period
that does not offer health
in health inswace
of tables 5U and 56 using a Mwent
for the .la$ ter thm
64 though
c=es
atilabitity,
coverage percentages
of the tr=sition
health insuance
insmmce
is higher for job qtits
costs is more damtic
couple
job sep=ations;
represents no job chmge,
is more fikely to de
the r-dts
of the health
pric-
trmsition
the probabfity
5C md 5d cofim
the probablfities
a tr-ition
whi~
h state health inswance
the sme
shocks to a st ate’s supply of private
Tabl=
for kvolmtmy
covwed
to hold.50
abfity
probability
this resdt
of the state poptiatim
irr state health insmace
COIU,
Table 5b shows the effect of&ages
Tables
in the proportion
rougtiy
sytiitric
of a tr=sition
from a job that offers health insWance
to the state health insmance
k
..1-
thee mexplf
and thm the restits reported h CO1-
Morco~er, tidividuak
who stay at thek s-e
shce it S an expensive proposition
tO the results in tables 5a - 5d. With a
cost measmes.52
Tabla
to one
5a ad
5b
ied. The coefficients reported h table 8a me ti m.nY
thee and k the other tables should be t~en
job may be hdated
for an employer to create . bedth &wmce
with a
&om market effects k the short
covmage benefit when none w-
pr.~iO=.ly OEer~.
51The CO,t of hosPital .m= h ~ stat= is a deteztiant
I me the average d~y
Sour.
hospital ro:m
&=g.
with
of tbe cost of health tisurmce
a state h cons tat
1983 doks.
pofiti.s
~e
sold fi that state.
prices =.
t&cn horn the
Book of E=lth Insumnce Data, Health hsurmcc Association of America, 1979-1992.
., The ~c=ption$ to th, ,t,t,d
relation~p
between state costs md hcdth bmance
43
transitiO.s appe= ~ cOIU=
show that the probability
insmance
coverage
~d
probabihty
is hcreastig
the avaage
hospital
The other
of loss of covaage
the &mges
ti state availabfity.
in both the avaage
room
ontstandng
Specification
The mgment
off=ed
feature of tables 6a – 6d is that the probabiEty
who leave thek job
job
where su&
ch=ge.
vidual’s
Authors
health
individuals
changing
of dsti~
inswance
gmeral,
coverage
insmace
who retin
in the
for those who qtit.
of job-lock
in whid
have
of a job
focussed
for health
wodd
ch~ge
criterion
or in the absence
on the detetinmts
insmace
coverage
for
of compensation
of a
of ~
To repeat their fd=
fi this foremst.
high demd
an individual’s
of changes in the utfity
in the event
as key factors
it is not possible
demmd
is a
Other factors,
thm
I estabhsh
jobs
to recover a measme
coverage,
factors
as an employment
to job-lo&
sud
because demd
wfich
in&-
arguat,
=e
detemed
cause at least a partial eti~tion
somce
measmes
the employer
of job-loti
using tidicators
ofly
do not have=
horn
of health
demad
period
of ~
effect.
effect
health ksmmce
An individud’s
in proportion
individud’s
u~biguous
cost of provitig
as state regtiat ions of health benefits -y
level of health
to its effect on employer
be more kportmt
contributors
factors.
these claims by deriving
over a specific
influmce
of job-lo&
benefit are what create a job-Io&
import~t
individual
2 of table 5a “md co]a tr-it
may occu
by their fem that a job &age
mobifity. 53 Also
on job
fi=ge
that health insm~ce
than for fidividtis
leave their job th~
of the W~ood
studies
dem~d
demand for health insm=ce
costs.
k
coverage.
fisumce
h
chmges
who have m wusua~y
jobs
incre~e
and Identification
is based on his best forecast
to h,
of health
room charge and the aual
up to this point impfies a model of job &anges
job mobifity
the state avtilabifity
Tables 5C and 5d show that the tr~ition
state hospital
job, and it is higher for those who involmtafly
Empitical
in both
charge.
wiU be lost is higher for individuals
sae
is dememtig
a
expression
for the probabfity
of time, which I discuss md =t@te.
1 of table 5b. k these ..1-s
ion from a job that offers hedt h ti”rance
kcremcs
that ~
individud
clak
.-
.
h state costs appem to decrease the chmce
covwage to one that does not. I attach the s-e
these receptions as to the exceptiom h tables 5a – 5d. I wdd
wi~
that the exception proves the de
&sd&er
of
to
that emerges
from these table,.
$, T&, ~ofit
wtid
is ,eParate fi.m
~e p-ti~y
deteded
the pokt
by a
that etittig
ti&vidutis
studies of j ob-lo& me mewmes of hedt h ksm~ce
preference for job mobtity.
44
dem~d
According
&screte
to the job
time interval,
whirh dete-e
Z = (u* - P~)/u~;
p.
fmction
Equation
nor~
is the mean
mployer
utihty offer distributio~
Equation
that 6r/8~.(t)
h,i(t) –
h
(19) =d”
(18) it is cle=
~ke~ood
whiti
utihty
jobs
dmhg
a fuction
of W the pm~eters
as:
nod
dasity
offer distributio~
deviation
fuctio~
au is the stmdmd
of the cmrent
of the curent
employer
utiEtY
tbt
for data on job &mges:
[r,(.)]”’ [1- r,(.)]l-M’
> 0 and 6r/6#v(t)
of elements
(21)
of (21) ae
< 0, which impfies
that r(.)
not identified.
is increastig
Frost, we
h
pi(t)
s
the me=
of
&,i(t) .54
words,
the probabihty
a pr=viowly
of a job
stated
&age
-d
states that h~e=es
is hcreashg
the me=
argument 6T/6pt(t)
has tra&tionaUy been adopted h job se=d
in the difference
of the cmrent employer
> 0 if 6ti.
(t)/6
p.(t)< 1. The --ption
between
utifity offer &tiibution.
that $u”(f)/JP.
(~) < 1.
models (e.g., Mortemen (1986)), b adopted here. The assumption
h the mem of the mmket offer &tribution
cause a 1.ss thm equal ticreme k the reservation
Utzty 1.”.1.
We&o
a
(18)
~ = (Y – Pu)/uU; #y is the mea
fuction
we can see that m=y
the market utifity offer distribution
54BY
i WW &mge
@ is the stadmd
Ug is the stmdud
L = fi
i=l
know
fuctiow
offer distribution
ad
is imp~citly
be rewrittm
of the m=ket
(2o) suggests the foUowing
By a short agwent
r;(t)
(10) c=
distribution
of the market utifity
from
that individud
= l\w*(t), At, A) = At A [1 - F(u;(t))],
for job &mges.
deviation
offer distribution.
the probabi~ty
At, A) = Pr(M~
U*(t) in (10).
where Z is the stmd=d
model,
At, is
ri(u”(t),
where M; is m indi=tor
se=ch
have resumed that 6z.(t)/8#Y
b the ~ected
(t) > u tti
utfity offer from the curent
w-
the as$mption
employer. Ttis =sumption
45
that the value of sc=d
hpfies
k sttictly time~tig
that 6r/6$y (t) <0.
This shodd
maket
come as no sqrise.
opporttities
me better thw
The term y;(t)
example,
is helpfd
8r~/6tii(t)
in m
individud’s
mtiket
equ~y,
cmrmt
a
fyi(t)
wage offer -y
when a
employer
on job mobfity,
-Ployer
dess
differential
employer
experience
md
ad
abi~ties
(whi&
chage
ket and expected
differenti~
educated
md
cmrent
selection
fli /6a~(i)
in the
offer includes
heath
medical
insuranm
comrage
coverage
the ~pected
at a high
ticreases,
but his
may decline
equa~y
po~cies.
compensation
offer from the muket
a
compensation
trmtment
from new health insuance
of both
compensation
is not identified
value of filtme
for health insmmce
cmrent
attainment
employer
offer from
the
will have no a priori effect
effect on muket
me both
~d
opporttities.
factors
wbifi
cmrent
arguably
Recall
the mguent
In that section I argued that the apected
the expected
compensation
An individud’s
offer from ? cmrent
cmrmt
abifiti=
a
individual
the opporttity
have a
from
the
compensation
employer
may exceed
by a given job me determined
should b: negatively
employer
when ~
the requirements
by the fire’s
to apply
related to the differential
compensation,
Hence, individuals
mpected
tidividuals
bcrease
of wage opporttities
The =pected
conditions
,teckology,
his fd
set of
because
att~ment
market and ~pected
cment
me more Ekely to work in a fim
to out~ow
of individud’s
betw~
~pected
mm-
the rate of ski~ acquisition
is highest
at
tend to outgrow jobs faster when they have fittle expe-
when they have a lot. Education
thus are less ~ely
The
medlcd
may not fiord
eqerjence
between
b
For
over time) to hls work.
low levels of experience.
rience th=
de-d
the skIUs reqtired
te&nology
Employment
zero.
for improved
his current job.
job, because
the fire’s
&pi(t) /6tii (t) may equal zero.
MS ~pectations
compensation
section above.
offer from the Inarket exceeds
of his cwrent
eqwl
and education
effect on mmket
‘outgrows’
employer.
cm be established.
ofle~ distribution
in~vidual
their ~pected
issues h a model of job mobfity.
a prior befief that the factor has a differmtial
opportmities
Employment
from their cmrent
of mderidmtification,
which is a detertinmt
md the expected
whm
may be mtiected.
individual’s
of pre-existing
facti any factoi
cment
curent
of market opporttities
due to the exclusion
h
whose
because
increase
exmple
/&tii(t) may dso
employer
price is dititished
~ectations
a prioii,
to change jobs
they mpect
the identification
md thus his job mobi~ty
because
offers from
in illustrating
relevant
are most fikdy
those whi&
is not identified
We have seen anothm
ptioti’,
hdividuds
a job th~
46
also be negatively
employer
whi&
individuals
with a propensity
shodd
compensation,
offers pot entid
mobfity
tito
to the
because
higher
for job advancement,
with lesser education
for job
related
jobs
atttiment.
that
do not
offer
health inswace
coverage
the =pression
permment
for r in (20)?
component
where pi represats
cmrent
pesstist
tiect
have pe-ent
=d
opportmdties.
- befieving
- always befieving
Pv,; ). ~ther
thoughout
the mdysis.
preference
How do preferences
for job mobifity
is represented
enter
h=e
as a
The int~retation
of
of p;(t).Suppose
employer
either an opttist
An individual’s
the time invmi=t
pi is that individuals
ad
hm been emphasized
way he tinges
of pref=ences
subjective
perceptions
A high mobiity
that his uket
of the ~ermtid
individud
opporttities
the opporttity
jObs Oft~
on job mobihw.
has a positive
due
of E;
=e always exceUent (high p.,;)
offered by his curent
because
between _ket
he is
– or a
employer is mmsuaUy poor (low
he is always ~satisfied
with his cwent
job (fdgh
p;).55
Failme
to control
for P{ in the esttition
Specifically,
it is know~
&mge
with above average fiequmcy
jobs
ployed in jobs
observable
that even titer
that offer low wages =d
&=acteristics,
tion includes
tismmce
the mmgkd
fdme
vmiable
effect on job mobfity
preferences
i’s cmrent
in the esttiation
=e
theoretical
comelated
mguent
charactetitics
estimates.se
bdividuds
1=s mobile individuals
coverage.57
i’s ctient
P woufd resdt
wage rate. The cuent
fied
employer
of a job mobfity
The esttiated
model
me included
on job mobifity
coverage
equation
job mobiity
effects.
of (20).
k the fitwatwe
of
appem to have a negative
of c, A, r, au, or Uv in the data.
wM*
equa-
esttite
even if health ins~mce
wage wodd
for
offers health
in a negative
measues
with the v~iables
to be em-
When m empirical job mobfity
on job mobfity
who
Thus, titer controMng
The same is true for the esttition
(21) with individual
I do not have any =pticit
vmiables
p=aetw
even if it had no true effect. I control for the effect of uobswved
by esti~tkg
thus otitted
insmance
to control for the mobservable
had no true effect on job mobfity.
tidividud
th=
for whether or not kdividud
coverage
in bi=ed
for observable
me more Wely
no he~th
effect of health insuance
which includes
contro~ng
resdt
the E@ p) > 0 and E(w p) <0.
m indicator
coverage
of (21) wotid
whid
parmetws
in the adysis.
associates
These
vmiables
me
wiff be biased if these
However,
WY of the o+tted
there is no
v=iables
.tlt ~ &o ~o~~ible that he is ~able to hold onto a job even if he wodd We to, because he frequently get. layed
off or k tied.
47
with the included
included
miables.
variables the par~eters
the otitted
variables
I approximate
tioml
Evm if the ontitted
r~(t)
atttimmt,
interactions
~e
of the fied
correlated
effects estbtion
wodd
with one or more of the
not be biased,. sfics.
nOne of
time varyhg.58
with a finear combination
state health
insmance
of measmes
costs, md
bet ween the health insmanm
late Ma&ia’s
variables wae
health
cost vmiables
strategy
for estimation
of a job-lock
fmction
k (.21) was esttiated
of employment
status.
ad
=perience,
I also include
the ikess
indicator
educa-
te~
for the
variable
to simu-
FoLowing
the logic
&ect.&g
Main Results
The Ekelihood
of the previom”
section,
that those individuals
not offer health imurmce
effects.
biased
Hetim
for ea@
person prescribed
in the malysis
in the sample,
estiwtes.
mobfity
h
the tied
‘*The
Ofitt.d
aticles
probit
me reported
cost measwes
coverage
It repr~ents
empirical
are ~lased downward
the fike~ood
as the equations stmda.d
.f Ma&ian,s
esttition
~OTh. fea~ibfity of the,, probits ,,qtitd
s-pie.
tidividual
model
sample
wiU be
ad
there me approximately
in the fist
is the &mge
between
W
seven
of observations
vmiables
per
used
for job mobihty
evidence
of a job ch~ge
Tfis
~
that the pmmeters
is increastig
of
of these
is evidence
has a l~ge
that
effect on
coverage
estimated
in
heterogmeity.
in the state supply of
ezrors k the probit cqUatiOn. The ~~e Of tfis
error is not identfied.
strategy wfich appeared ember k the ht=ature
that I restrict the s-Pi.
The cost w= m ticrewe
two.,
.of health hs~ance
by the effect of mobs:rved
of table
in the estimates
one Wd
job mobifity.
and determinants
the estiated
two COI-S
that occms
COIDS
in m equation
,Wiables wm, h.weve,, ..nse h.r.ase
6ss,= the &,cu,,ion
a data
which is close to the nuber
to health insmance coverage
effects estimates
tiefficiency is ufio-
from the l~g~
In my smple
signs when I control for indvidud
equation.
job-lock
individual@
that do
to job mobifity.
of health inswance
est fiat es of coefficients
effects probit
Also, a case cm be made that au of the explanatory
insmance
fiange
with md without
in the &ed
oyer time for mch
of the core job mobifity
to health
the endogeneity
in a job
estimates
that stands out in these COIWS
the coefficients
pmmeter
individual
here are exogenous
The featue
model was estimated
me not strictly ~ogenom.
by Heckman.
The estimates
etisting
shows that parameter
variables
equations.eo
who are most motille are selected. into jobs
each probit
if there me too few obsermtio.ns
observations
9a.
coverage,
(1981)
the expl~atory
in a series of probit
h esttiated
48
to 800 k&viduak,
st~dmd
errors,
review.
who w=re r=dO~y
selected
.D
health
immmce
coverage.
medicd
coverage
It is deer-kg
me.
indvidud
k
si~
fim
effects =e
that individu&
at a fast rate) md IUS qti&ly
p=-etm
si~
to =pen.m
but this negative
=d
(1994).
a deterrent
table
(2)-(4)
Thwe
According
jobs
Accortig
is pinged
coverage
on
of the effect of
of years of employmmt
Mthough
this sign change does not
pr=mted
when they me yomg
to the tied
~eri-
tiects
hwe.
I tied
(when
SMS
more slowly).
esttiat-
me
The
of the coeffi-
change jobs less frequently at tigher levels of experience,
interactions
tends to dissipate.
betwem
were included
=
a reflection
of an ties.
ad
th=
Thee
pre-etistbg
medi-1
for those who =e healthy.
of the fom esttited
the health
of the sp edifications
effect of the state cost measmes
who me si~
9a suggests that this may be true.
the occmence
if m individual’s
then the det-ent
be stronger for in~viduals
qtickly
on job mobifity
te-
the squ=e
by the theOry Of jOb mobifity
t? her =guent,
to job mob~ty,
the Chmge in the state price of
as they age (when SWS =e hcre~tig
kdividuds
include
mst measmes.
Ma&i=
shodd
e~ers~
ad
added to the model.
hems out this .lti.
effect of experience
Specifimtions
ksmace
to ezpers~,
out~ow
kcreastig
ci~ts
me
of health insm=ce
effects.
cOnclusiOn, it was pre~lcted
&mge
k the state supply
by theory, cndy emerges when the modd
that the coefficient
whm tied
the last section
cbge
of the effect of the state cost of health insumce
specific job mobihty
It is dso interesthg
ence, chages
amual
in the state price of medicd
which is predicted
wobswved
~Y
the percent
However, this pictwe
job mobihty,
fiPIY
=d
used by
conditions
me
on job mob~ty
The last colm
hteraction
of
term coefficients
me negative.
The co~cients
to illswll do not appe=
Esttites
of the coticiat
k futher
evidence
dewd
to ikess
than a deterrent
effect
coverage.
status is positive.
caot
be identfied
Nonetheless,
K mything
The v=iation
ustig
in these estfiates
mmswes
of m inti~ldud’s
h W but one of the CO1-S
ihess
in table 9a.
the esttited
is a cause of inmeased job moblfity
rather
to it.
Results
Tables 9b-e repeat the probit adysis
specifications
of the s-pie.
femles
range from -.756 to 6.013.
that a job-lock
for health insmace
coefficient
Other
to Mess
to be robust amoss the model specifications
mdy.
test the robustness
Table 9b restricts
The des
ody
ad
w~e
restricting
of the tin
the smple
fedes
the smple
results to +ages
to males ody,
ody
smples
49
md
by sex ad
uitd
status.
ti the demograpMc
table 9c restricts
generate sM=
resdts.
These
imposition
the s=ple
One si@=t
to
difference
is for the esttiated
Gains in job experience
coefficient
tti”jol
have a much stronger
experience
negative
.- –.087
for roles
effect on job
and – .013 for fedes.
mobfity
for males compared
to females.
AH the esttiated
of the estfiated
coefficients
coefficient
a negative
but statistica~y
eqtivdent
coefficient
comtsfor
b
insignificant
the fetie
saple”is
and the coefficient
to —ried
respondents
compared
co~cients
tables 9a and 9b. hcreases
hsi~ficant
of the co~cient
sigtific~t.
saple
wMe
explmation
negative
ad
generates
to pet-ch,
NO simple
c~efficient oflnhospis
of pctchghp is negative
atttimmt
to mmried
smple
ks
a stronger
respondents.
ody
the
ac-
=dsig@cat
sigtificat
for md~
restricted
for mmried
negative
These resdts
the average age of respondents
of the health insmmce
s~ples,
individuals
tid
respeccompared
effect on job mobi~ty
for
may be coutio~ded
by
is higher thm
the average in
whifi
for both.mmried
has a negatively
k estimated
cost variables repeat the pattern
h the state supply of health insw~ce
of coverage increase job mobfity
coeffiti~t
ad
ody
sample.
The estkated
is for pet-ch
positive
The m~es
from married only and mmmried
education~
the fact that in the mmried
the married
esttiate
has a much stronger effect on job mobifity
individuds;
mmmried
sample.
sign with the exception
for females.
Tables 9d md 9e report rwtits
tively. Job =perience
fixed &ect
However, theestkated
fedes,
negative but insignificant
have. the predicted
of pet-ch in the males ody
this contradction.
for both males ad
of the mst variables
md married
signed coefficient
to be – .124 with a stadard
one.
50
in the fied
coverage
observed
or decreases
indlviduds.
effects col~
The ody
in
in the cost
exception
of table 9e. The
error of .726, so the aberration
is clemly
m
Conclusion
The =pent
presented
relatiomtip
between
view of the sodd
of job
ch=ges
medicd
=d
condititi
evidence
job chmges
=d
consequences
&mges
d-ges
of employ=
k insmmce
modd
for a more complete
in health
Eded
pofities
ch=ges
insmmce
to judge,
of the theoretid
nd
for a broad=ed
coverage.
Previous
on the exclusion
u a potential
in imm=ce
exposition
coverage
have focussed
by those stuties has been &fficdt
of expected
insmmce
hdth
coverage
horn new health insw=ce
presmted
a complete
in this study has mWed
studies
of pre-etisting
deterrent to job &snges.
The
because no study has evsr provided
coverage
md
th~
role irr detetining
job
mobfity.
This study has attempted
ties
=pfitit
sation
horn
conditiom
=e
the value of h=lth
fl
=e
potential
a potential
sowce
job
dso
coverage,
supports
devdoped
costs of provitig
demd
The empirical resdts
to remedy
the negative
health inswace
of employers
medicd
health
condltims
insmance
in health irrsmace
have importmt
sho~s
medid
mdy when
to the maket
medicd
they
supply
conditions,
to employees
coverage ad
POUCYimp~catiom
dtimately
coverage
across employers
portable
to cacel
thek
health
for health insurace
~ed
employee
coverage.
WN lose his health insmmce
etinates
are, at best, a smW
coverage
effect of employer
who offer health insmace
strengthen
inme=ing
that preetisthg
of
cm have
tested using data from the NLSY. The emptilc~
hmlth
insmmce
the deterrent
to job
mobfity,
coverage
shotid
det ement to
were fowd
to
on job mobfity
coverage,
either through
as an employment
thus actua~y
rdorm.
on job
be cmeftiy
insmmce
Their effect -y
coverage,
coverage
inswmce
health
coverage by rtising the employer cost of health insuance
codd
levels of compen-
effsrt is sigfifiat
to pre-etisting
whi~
factors.
may cause some employers
employer
of job mobfity
of futwe
clew
that negative
in addition
the view that pre-efisting
Employer
do individud
hdividud
mkes
but that thek
here w=
have a much greater effect on &ages
m&tig
model
imphes
modd
md expectation
dynamic
of job-lo&,
a dyntic
eflect on job mobfity.
mobfity.
thm
coverage
The
The model
The model of job mobfity
evidence
insmace
Hnked health insmmce
a deleterious
that gap by expositing
employers.
severe and prolonged.
employer
to fl
Propos&
mobfity
evaluated.
benefit plms
by
They
or to rstrict
thus reducing the overd
nmnber
be to increase the risk that m
a job
benefit.
reducbg
chmge
or if his cwrent
k&ectly,
the proposal
job mobtity
rather
thm
conditions
horn
it.
PrOpOsds to prohibit
insur-ce
companies
from excluding
51
pre-etisting
medicd
new health insm~ce
conditions
tight
to tid
a new job
(induting
that offers health insumce
cost of offetig
health inswmce
benefit pl~s
that c~ently
p ofiti~
efists
effects.
that medical
pofiq
kdividuds
those with pre-dsttig
coverage,
god
were not fi&ed
52
ad
the coverage
of increaskg
to employers
between health tismmce
jobs,
c~verage
medicd
renditions)
a positive
wodd
effect
may be less able
since the legal prohibition
covwage,
or reduce
with costly
coverage for their conditions
if they were to &mge
health insmmce
here suggests that the sodd
etist if health insmace
the fid
have tintended
by the expectation
W individuals
increase the employer
The =dysis
dso
from a new health insmance
However,
their employee
my
be encomaged
not be excluded
for them.
poEties
wodd
came some employers
tmd
to
to cacel
in those plans.
job mobifity
may ody
to a level that wotid
be atttined
and mployers
by removing
in the U.S.