MAX LIBRARIES - DEWEY
Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium IVIember Libraries
http://www.archive.org/details/savingbabieseffiOOcurr
Dewey
HB31
.M415
'0
^
,
working paper
department
of economics
SAVING BABIES: THE EFFICACY AND COST
OF RECENT EXPANSIONS OF MEDICAID
ELIGIBILITY FOR PREGNANT WOMEN
Janet Currie
Jonathan Gruber
94-11
Dec.
1993
massachusetts
institute of
technology
50 memorial drive
Cambridge, mass. 02139
SAVING BABIES:
THE EFFICACY AND COST
OF RECENT EXPANSIONS OF MEDICAID
ELIGIBILITY FOR PREGNANT WOMEN
Janet Currie
Jonathan Gruber
94-11
Dec.
1993
MTtUBRSBiES
JUL
1 2 1994
RECEIVED
Saving Babies: The Efficacy and Cost of Recent
Expansions of Medicaid Eligibility for Pregnant Women
Janet Currie
UCLA
and NfBER
Jonathan Gruber
MIT
and
NBER
December, 1993
We
Diamond, Sarah Feldman, Guido Imbens, Alan Knieger, Jim Poterba, Jon
Stem, Duncan Thomas, Aaron Yeiowitz, and to seminar participants
at Cornell, UVA, and the NBER for helpful discussions, to Sharieff Mansour and Nancy Cole for
excellent research assistance, and to Marilyn Ellwood, Michael Fischer, Ian Hill, and Aaron
Yeiowitz for providing information on Medicaid policies. Janet Currie gratefully acknowledges
support from the National Science Foundation (SES-9 122640) and the Alfred P. Sloan Foundation.
The views expressed are solely the authors' and should not be attributed to either organization.
are grateful to Peter
Skinner,
Doug
Staiger, Steve
Abstract
A
key question for health care reform
in the U.S. is whether expanded health insurance
improvenwnts in health outconoes. We address this question in the context of
dramatic expansions in the Medicaid eligibility for pregnant women that took place during the 1980s.
We build a detailed simulation model of each state's Medicaid policy during the 1979-1990 period,
and use this model to estimate 1) the effea of changes in the rules on the eligibility of pregnant
women for Medicaid, and 2) the effect of Medicaid eligibility changes on birth outcomes in aggregate
eligibility will lead to
Viial Siaiisncs data.
We
eligibility
have three main findings.
First, the
of pregnant women, but did so
expansions did dramatically increase the Medicaid
at quite differential rates
expansions lowered the incidence of infant mortality and low birthweight;
percentage point increase
in infant mortality
of
in eligibility
7%. Third,
among 15-44 year
earlier, targeted
old
changes
relaxations of the family structure requirements from the
on
birth
levels.
outcomes than broader expansions of eligibility
We
women was
in
Medicaid
AFDC
to all
who became
eligible
we
estimate that the 20
associated with a decrease
eligibility,
compared
much lower takeup of Medicaid
under the broader expansions.
to conventional estimates
We
of the value of a
life.
making insurance policy
effective.
find that the
life
We
insurance expansions can improve health, but that translating eligibility to coverage
link in
effects
with somewhat higher income
targeted expansions, which raised Medicaid expenditures by $1.7 million per infant
fairly cost effective
such as through
program, had much larger
women
suggest that the source of this difference was the
coverage by individuals
Second, the
across the states.
saved, were
conclude that
may be
the key
Will the extension of health insurance to the uninsured improve their health?
question underlying the current debate over health care reform.
be a necessary precondition for improvements
eligibility for
may
insurance
insurance coverage.
improvements
It
is
This
is
a key
Although insurance coverage may
in the utilization
of medical care, expansions
in
not translate into increased utilization of care, or even mto increases in
possible that
also
increased utilization of care will
not result
in
in health.
This paper sheds light on these issues by investigating the relationship between the health of
newborns and recent expansions of the insurance coverage provided
Medicaid program. With 10 infant deaths per 1000
births, the
to
pregnant
women
U.S. infant mortality rate exceeds that
of 20 other industrialized nations (U.S. House of Representatives, 1992). This high rate
to reflect large
numbers of unhealthy newborns.
improves neonatal health, there
is
Hence,
scope for a decrease
under the
to the extent that
in this rate
is
thought
adequate prenatal care
through the promotion of prenatal
care (Institute of Medicine, 1985).
In an effort to increase die use of prenatal care, the past decade has seen a rapid expansion
in the eligibility
of pregnant
women
for Medicaid, a federal-state matching entitlement
provides health insurance for the poor.
receipt of cash welfare
eligibility
Medicaid had traditionally been
eligibility to
was extended
to a
very low income
number of
women
in single
additional groups of pregnant
women
This experience provides a case study of whether expanding health insurance
actually
improve infant health, or whether poor
beyond the reach of government insurance
during the
eligibility
can
infant health in the U.S. simply reflects other factors
policy.
study the effert of this government intervention,
changes occurred
tied to the
parent households.
1980s.
To
that
payments under the Aid for Families with Dependent Children (AFDC)
program. This generally limited
However,
Eligibility for
program
at different rates across the states
in
we
exploit the
faa
that the eligibility
order to identify their effects on birth
outcomes, birth inputs, Medicaid coverage, and Medicaid expenditures.
analysis
is
a detailed simulation model of each state's Medicaid eligibility rules for pregnant
We
over the 1979-90 period.
in
women
apply this model to data from the Current Population Surveys (CPS)
order to quantify the effects of changes
We
The baclcbone of our
in the rules
on
eligibility
and on actual Medicaid coverage.
then use aggregate Vital Statistics data to examine the effea of the expansions on two widely
used indicators of infant health: the incidence of low birthweight, and infant mortality.
Using these
estimates in conjunction with data on Medicaid expenditures from the Health Care Financing
Administration
(HCFA), we then examine
the costliness of Medicaid expansions.
information on the use of medical services by pregnant
women from
the fraction of
increase of
changes
women
major fmdings.
140%. This increase occurred
a lower incidence of
was stronger than
the effect
in
Medicaid
and a decrease
on low birthweight.
eligibility
who had been
Third,
program, and expansions of Medicaid
ineligible for
AFDC
extensions of Medicaid coverage to
of the federal poverty
Medicaid
all
level).
women
on
eligibility
two categories.
infant
changes
"Targeted
in eligibility criteria for
eligibility to
for reasons of family structure.
all
were associated with
in infant mortality; the effect
In particular, the changes can be divided into
AFDC
34%, an
to
across the states, causing dramatic
expansions' were narrowly targeted changes which included changes
welfare under the
inputs.
of pregnancy from 14.2
at quite different rates
births
use
estimate that the expansions of the 1980s increased
Second, increases
generosity.
low birthweight
are not created equal.
we
eligible for public insurance in the event
in relative state
mortality
First,
we
the National Longitudinal
Survey of Youth (NLSY) to ask how the policy changes affected the use of birth
We have three
Finally,
cash
demographic groups
"Broad expansions" were
with incomes less than specified levels (e.g.
185%
These two types of policy changes affected quite different populations,
suggesting the potential for heterogeneity in their effects.
In fact,
we
outcomes, while the broad expansions had
is
little
in the differential effects that these policies
much lower takeup
featured
had sizeable and significant effects on birth
find that the targeted expansions
rates,
We
effect.
suggest that the source of this difference
had on Medicaid coverage; the broader expansions
even among otherwise uninsured mothers.
The
targeted
expansions were associated with a Medicaid expenditure increase of $1.7 million per infant
saved.
This figure
is
fairly
low compared
to recent estimates of the value of a
high relative to clinical estimates of the costs of saving a
life.
However,
through improved prenatal care.
life
are unable to resolve whether this reflects a failure of prenatal care to deliver
its
life
it
is
We
promised savings,
or a failure of mothers to obtain such care once they were eligible for Medicaid.
The
rest
of the paper
is
about our measures of newborn health.
effects
on
eligibility.
Part
In Part
laid out as follows:
In Part
II,
we
I,
we
provide background information
discuss the Medicaid expansions, and their
investigates the effect of these expansions
III
on
birth outcomes.
Part IV
discusses the differential effects of the targeted and broad expansions on Medicaid coverage.
V
investigates the cost-effectiveness of the targeted expansions using information
expenditures and individual use of prenatal care.
Part
Part
on Medicaid
VI concludes with a discussion of
the policy
implications of our findings.
I.
The
infant nwrtality rate and the incidence of
examined indicators of
1980s.
less
Background oo Birth Outcomes
infant health.
The incidence of low
in the
U.S.
low birthweight are two of the most frequently
Figures la and lb plot the trends in these measures over the
birthweight, defined as the
number of
live births per
1000 weighing
than 2500 grams (approximately 5.5 pounds), declined from 68.7 in 1979 to 66.6
then rose to 69.4 by 1990.
in
1984, but
In contrast, infant mortality declined throughout the decade, although
it
declined at a slower rate between 1984 and 1989 than
underscore the
fact that although they are related,
had
it
earlier.
These differing trends
low birthweight and infant mortality measure
different aspects of birth outcomes.
Low
birthweight
of low birthweight are
is
at
a key indicator of the underlying health of the fetus.
Horbar
et al.
who
among normal
(1993) found that
between 601 and 1300 grams
at birth,
a decrease of approximately
10%
birthweight infants (Office of Technology Assessment,
in
a sample of very low birthweight children weighing
each increase
in the probability
in
birthweight of 100 grams was associated with
of death, other things being equal.
In contrast, infant mortality rates reflect not only the health of the fetus as
measured by
New
birthweight, but also the effect of any interventions that occur during or shortly after birth.
technologies have had dramatic effects on the survival rate of low birthweight infants.
al.
are
high risk of neonatal mortality and experience post-neonatal mortality rates
10 to 15 times those found
1987a).
Children
Buehler
(1985) report that improvements in birthweight-specific mortality rates accounted for
91%
et
of the
overall decline in neonatal mortality between 1960 and 1980.'
These interventions, however, are often very expensive.
although babies weighing less than 25(X) grams account for only
they account for
57%
of the cost of neonatal hospital care.
9%
Schwartz (1989) reports
that
of neonatal hospital caseloads,
The average
cost of caring for a
surviving low birthweight baby was $9,712 compared to $678 for an infant weighing more than
2,500 grams.
These costs
rise as birthweight falls; in
birthweight below 1000 grams was $118,000
'
More
recently,
Horbar
(OTA, 1987b). Moreover, survivors
et al. (1993) report that as
mortality reported between 1989 and 1990
1984, the cost of saving an infant with
nwy
much
as half of the decline
are at high risk
m
national infant
be attributable to the introduction of a
new
therapy
replacing an essential substance in the lung (pulmonary surfactant) that is not
manufactured by the fetus in significant quantities until the 33rd week. This therapy was introduced
for artificially
in
October 1989.
^
of handicaps such as cerebral paJsy of significant degree, major seizure disorders, blindness,
(McCormick
deafness, and learning disorders
ei al.,
OTA,
1992;
1987b, Chaikind and Corman,
1990).
The high
cost of caring for low birthweight infants and their uncertain future should they
survive, have led policy makers to emphasize the prevention of low birthweight through the
There are a number of ways
promotion of appropriate prenatal care.
improve
For example, approximately two-thirds of
fetal health.
term, and Creasy et
al.
(1980) found that over
inexpensive ($10 to $20) screenings
60%
all
that early prenatal care can
low-birthweight births are pre-
of these cases could have been identified using
in the first prenatal care visit.
Several clinical studies cited in
the Institute of Medicine's influential 1985 report suggest that providing appropriate prenatal care
to
women
identified
by these screenings
a cost of between $400. and $500. per
(at
reduce the incidence of low birthweight of more than 20%.
saving a
life
through improved prenatal care
As has been noted by
outcomes among
most $142,000 per
number of economists, however,
a
women who do
is at
and do not receive prenatal care are
Compared
see Harris (1982) for an extensive discussion.
This figure
cost $20,000.
is
calculated as follows.
One would
Of
these,
Prenatal care for these 28
of low birthweight by
much
that the cost of
saved.
studies based
likely to
on differences
based on survey
smaller effects of prenatal care on
screen 1(XX) pregnancies at a cost of $20 each would
To
20%, or
47
will
women
six babies.
in
be biased by selection;
expect 70 of these pregnancies to result in low birthweight births
absence of any intervention.
screening.
life
to clinical studies, studies
data that attempt to control for this selection typically find
'
These figures imply
woman) could
be pre-term, and 28 (60%)
will then cost
Using the Vual
will be caught
in the
by
this
$14,000, and will reduce the incidence
Statistics data
below,
we
estimate that
a decrease of one low birthweight baby per 1000 births lowers the infant mortality rate by 0.04
deaths per 1000 births. Thus, the $34,000 spent to reduce low birthweight births by six babies will
save 0.24 lives, for a cost of $141,667 per
benefits of prenatal care, since
some of
life
saved.
the babies
higher birthweight, will be less impaired later
who
in life.
an understatement of the
would not have died, but who are now of
This figure
is
birthweight (Rosenzweig and Schultz, 1982, 1983, 1988; Frank etaL, 1991;
Grossman and Joyce,
studies focus
These differences
1990).
on the gains
that could be attained
die impact of prenatal care as
it is
under
practiced in the field.
all
pregnant
In
women,
summary,
For example,
women
the cost of saving an infant
life
et al.,
1987;
also reflect the fact that clinical
ideal circumstances,
reduce pre-term delivery were delivered not only to the
but to
may
in results
Gorman
if
whereas surveys reflea
prenatal care designed to
identified as high risk
would
rise to
by screenings,
over $1.3 million.'
the available clinical evidence suggests that while both reduaions in the
incidence of low birthweight and high-tech neonatal care can reduce infant mortality rates, the former
is
the
more
cost effeaive policy.
the use of prenatal care
coverage of pregnant
was
woman
in
Decreasing the incidence of low birthweight through increases
faa the primary motivation
The Medicaid Expansions
Background*
Historically,
in
Medicaid
that took place during the 1980s.
II.
a)
for the expansions in the
in
AFDC.
Medicaid
This linkage with
the existence of the
eligibility for
AFDC
restricted access to the
AFDC-Unemployed
which the primary earner
is
women and children has been closely tied
program
in three
ways.
to participation
First, despite
Parents program which provides benefits to households
unemployed,
AFDC
in
benefits are generally restricted to female-headed
'That is, there would be $500,000 spent in delivering prenatal care to all 1000 women in the
sample, and 9.4 pre-term low birthweight births would be prevented (since 100% of these cases
would now be caught). This implies a cost per life saved of 1.33 million.
which Federal matching funds were
available. States can cover additional groups under Medicaid, but only if they pay the full cost of
coverage. We do not have mformation about these "state-only" policies. In addition, we only
There were also expansions of the Medicaid
discuss policies that applied to pregnant women.
*
The
policies discussed in this section are those for
coverage provided to children
in this era; see
Yelowitz (1993) for
details.
households.* Second, income cutoffs for cash welfare vary across states, and can be very low.
example.
In
Texas, the cutoff for a family of 4
the stigma of applying
for cash welfare
in
24%
1979 was only
of the poverty
programs may have prevented
line.
Third,
from
families
eligible
For
receiving Medicaid benefits (Moffitt, 1992).
However, from
the inception of the Medicaid
program,
extending Medicaid benefits to some groups of pregnant
options expanded rapidly during 1980s, in a manner that
eligibility
changes can be divided into two types.
defmed groups who had not
women
whose income was somewhat above
Also included
Medicaid
in this
category are changes
eligibility.
Because these
in
first
AFDC.
AFDC.
This included
AFDC
AfTX) income
These
In brief, these
type expanded coverage to narrowly
diese cutoffs but
eligibility
not on
detailed in the Appendix.
and two parent families whose income was below
individuals
in
traditionally qualified for
have had the option of
women who were
is
The
states
first
time pregnant
cutoffs,
and "Medically Needy"
who had
large medical expenses.
limits,
which carry with them changes
changes are narrowly targeted,
we
label
them
"targeted expansions".
Beginning
States
were
income
first
in
April 1987, the income requirements for Medicaid were also greatly liberalized.
given the option and then required to provide Medicaid coverage to
levels that greatly
exceeded the
AFDC
uniform minimum threshold had been established
with incomes up to
incomes up
to
185% of
income requirements
*
Not every
133% of
state
- all
limits in
states
most
states.
were required
to
By
the poverty line.
In
what follows, we
with
April 1990, a
cover pregnant
the poverty line, and stales had the option of covering
women
women
with
will denote these relaxations of the
as "broad expansions".
AFDC-UP program over our sample period, and eligibility requirements
of 1990 only 5% of the AFDC caseload qualified under this program (U.S.
had an
strirt. As a result, as
House of Representatives, 1992).
are
income
women
b) Effects
on EUgibiUty
In order to estimate the effects of these eligibility expansions,
we have
built a detailed
simulation of Medicaid policy in 49 states and the Distria of Columbia, over the 1979-1990 period.*
The construaion of
12 years of
CPS
simulation
this
is
described in detail
from 14.2%
In this seaion,
Medicaid coverage
to
recession years,
34%
of
this
in
women
in
eligibility criteria
figure indicates that
population between 1979 and 1990.
became
sample
when
moderate increase
eligible
because they
The percent
However, the
stricter in the early
is
restricted to
fell into
for
that the
eligible rose
eligibility increases
During the
poverty, so eligibility increased
years of the Reagan adnunistration.
women
with incomes below twice the poverty
to the
AFDC
in eligibility associated
program
in the early
most dramatic changes
'80s.
line.
fell
Both figures show a
with increases in the coverage of unborn children and two-
earner families mandated in the Deficit Reduction Act of 1984
*We
is
have been
business cycle effects are controlled for in this way, eligibility
1982 as a result of cutbacks
show
The CPS
CPS who would
in the
each year had they become pregnant.
many women became
In Figure 2b, the
The
use
demographic and income information
of the early 80s show that these estimates are sensitive to business cycle effects.
even as
we
this simulation.
Figure 2a shows the fraaion of 15 to 44 year old
eligible for
Appendix.
data to simulate the effea of the expansions on Medicaid eligibility.
the largest available annual data source with the requisite
undertaking
in the
in the
number of
exclude Arizona from the analysis because
program.
8
(DEFRA
eligibles
it
'84).
But the figures clearly
were associated with the relaxation
does not have a conventional Medicaid
of income restrictions
80s
in the late
-- eligibility
The aggregate time trends shown
growth of
On
Ohio saw steady
On
between 1988 and 1990.
1988, but large jumps in the
This variation
women
eligible for
we examine
eligibility.
states,
is
rises in eligibility
show two
from 1979
explored further
in
variability in the
with very different patterns:
to 1988,
and a relatively small jump
little
change
in eligibility until
each
in
Table
state, in
1.
Here,
we show
the fraaion of 15-44 year old
each of the years 1979, 1986, and 1990.
For 1990,
under the targeted and broad expansions separately, as well as overall
eligibility
The growth from 1979
a reduction in states such as
is
states
1990.'
two years of the sample.
to
1986 under the targeted expansions
such as Colorado, Minnesota, and Virginia, while there
but there
mask considerable
the other hand, Florida experienced
last
Medicaid
Figures 2a and 2b
Figures 3a and 3b
eligibility across states.
the one hand,
in
100% between 1987 and
increased almost
New
is little
The growth from 1986
Jersey.
to
also substantial variation; eligibility growth in Mississippi
is
very dramatic
growth
in
1990
positive for
is
is
some
in
Delaware and even
all states,
over three times that of
New
Jersey.*
It is
this substantial
for our study.
change within
If states are
and over time that provides the identifying variation
ranked by the frartion of the 15-44 year old female population that
eligible, the correlation
between rank
experienced a change
ranking of
in
states
in
1990 and rank
at least
10 positions:
in
1979
is
Illinois fell
only 0.33.
Twenty-seven
is
states
from the 8th most generous
state
^Within the set of targeted expansions, the policies that had the biggest effect were those that
covered first-time pregnant women. The faa that these changes, and not the coverage of two-parent
families, had the largest impact reflects the fact that few married couples were poor enough to
qualify for the Medicaid
program under the
AFDC
income
cutoffs.
changes were "expansions"; half of the states
increased eligibility from 1986-1990, and half of them reduced it. On the other hand, almost every
state shows a sizeable increase in eligibility under the broad expansions.
•
Clearly, not
all
of the targeted
eligibility
to the
38th most generous, while Mississippi rose from the 44th most generous state to the state that
makes
c)
the highest fraaion of
Targeted
vs.
population eligible.
Broad Expansions
Throughout
expansions.
its
Our
this
we
paper,
will
rationale for doing so
is
distinguish between the effects of the early and broad
presented
in
Table 2, which highlights the heterogeneity
of the populations affected by these two types of expansions.
covered under the targeted expansions, while the third looks
under the targeted expansions, but were
are based
eligible
at individuals
column shows
the
the set of individuals
who would
not have been
under the broad expansions.
All
columns
on data from the 1990 CPS.
While individuals covered by
relative to the
fiill
somewhat
either the targeted or broad expansions appear disadvantaged
sample, the broad group appears to be nx>re similar to the sample average than
to the targeted group.
is
first
The second column examines
characteristics of the entire eligible sample.
eligible
The
Compared
wealthier,
more
to the targeted group, the
likely
to
Furthermore, the broad-expansion group
group covered by the broad expansions
be white and married, and slightly older on average.
is
much more
likely to
be working, and
less likely to
be
receiving any form of public assistance than the group covered by the targeted expansions; in fact,
the group covered by the broad expansions
average 15-44 year old female.
is
less likely to
While the two groups are equally
composition of insurance coverage differs substantially.
expansion group
7%
is
be receiving public assistance than the
covered by Medicaid, while only
group does.
10
be uninsured, the
Thirty-eight percent of the targeted-
13% of the broad-expansion group
of the targeted group has employer-provided health insurance
die broad
likely to
in their
own name,
is;
and only
while
25%
of
This population heterogeneity suggests that the two types of expansions may have had very
different effects
on
birth
outcomes.
We
will allow for such a differential
effea
in the
estimation
below.
III. Eligibility
a)
and Birth Outcomes
Methodology
We
from
examine the
Vital Statistics,
effect of the eligibility expansions
the
that
two measures of
1979-1990 period. The
is,
we
outcomes using aggregate data
birth
which reports the incidence of low birthweight
infant mortality in each state
this data to
on
and year for the
eligibility
first is
among
full
sample of births
15-44 year old
all
the actual percentage of
use the data presented in Table
1
than 2500 grams) and
(less
in that state/year.'
women
women made
for each of these years,
in
each
state
eligible
matched
We
match
and year over
under expansions;
to the birth
outcomes
data for that state/year.
A
may
potential
drawback
to this strategy,
however,
is
that the
CPS
fraaion-eligible measures
incorporate business cycle effects, or other shocks specific to states and years.
showed, for example,
that the recession
of 1982 was associated with increases
despite the adoption of stricter eligibility criteria.
in
Medicaid
To
'
eligibility
Similarly, the fact that in 1990 Mississippi had
the highest fraction eligible of any state reflects both the generosity of the state
relative poverty
Figure 2a
program and
the
of Mississippians.
the extent that relevant state and year specific characteristics are not captured by state and
explored the effea of the expansions on low birthweight in the NLSY
While the findings were supportive of our conclusions using aggregate data,
In an earlier draft,
data discussed below.
the confidence intervals
we
were too large
to
draw any
unable to investigate infant mortality with the
NLSY
II
useful conclusions.
due
to small
sample
Furthermore,
sizes.
we were
year dummies,
coefficient
on the
a state recession
frartion eligible will be biased
between Medicaid
In order to
We
first
source of variation
this
Suppose, for example, that
overcome
this
fraaion of this same sample of
By using
in eligibility
could induce a spurious positive correlation
and low birthweight. '°
eligibility
problem,
took a 1-in-lO sample of our
that year.
by omitted variables.
associated both with increases in eligibility and with a higher incidence of low
is
Then
birthweight.
they are not constant within a state, or across states within a year), the
(i.e.
CPS
we have
created a measure of "simulated eligibility".
women
cohort of
women who would have
for each year.
in
Furthermore,
we
derives from having relatively small cells for
These models
some evidence which suggests
states in the
Medicaid policy as exogenous
treat state
that this
assumption
is
not true.
were more
and outreach programs.
likely to
In the
'°A similar bias could arise
level
find for
example
in
to birth
indicates that
Simulated
women
is
in
much
our estimate that
in
Gold
is
year dummies included
in fact
et al. (1993) report that states
women who
delayed obtaining
if
adverse trends
will include state fixed effects
eligibility is
will control
outcomes are correlated with the absolute
largest impact where the population is poorest.
in birth
much
fiaiter,
which indicates
better than those in other states.
flatter
which
On
of women
that
fell
sharply
on average, young
the other hand, the increase in
than the increase in actual eligibility in Mississippi, which
Mississippi were poorer than a nationally representative group of
eligibility will still
we
independent of other
outcomes; there
that in Connecticut, actual eligibility for this cohort
Conneaicut fared
simulated eligibility
in the
models below, we
Simulated
in the mid-to-late-80s.
women
state in
adopt optional expansions and to set up complementary advertising
of poverty, since expansions may have their
"We
each
CPS.
with high proportions of low birthweight births and high fractions of
prenatal care
is
reduce the sampling variability
some
in
each state simulation,
obtain an estimate of the fraaion eligible that depends only on state rules, and
characteristics of states."
then calculated the
been eligible for Medicaid
same group of approximately 1000 women
the
We
vary with aggregate cyclical conditions, but this variation
in the regression.
12
is
women.
absorbed
for such spurious correlation
between time invariant
state characteristics
and Medicaid policy.
b) Overall Results
Table 3 presents the basic regression
All regressions include a full set of year
results.
effects, to control for the strong trends in infant mortality
They
la and lb.
across states that
is
identified
may be
correlated with both Medicaid policy and birth outcomes.
by the deviation of Medicaid
When we
birthweight, but
illustrated in Figures
also include a full set of state effects, to control for time invariant differences
The top panel of Table 3 models
1980s.
and low birthweight
eligibility
is
we
fraaion eligible, and becomes significant
lead to a redurtion of
in eligibility
2%
in the rate
state-specific
find a negative effect
The estimate
not statistically significant.
percentage point increase
its
mean.
the overall effects of changes in Medicaid eligibility in the
use the actual fraction eligible,
it
from
at the
5%
level.
rises
The
on the incidence of low
when we use
the other hand, there
is
7%
We
of low birthweight.
conclude that there
eligibility
eligibility
is
is
some evidence
relatively small.
a sizeable and significant effect of expanding Medicaid eligibility
eligibility used.
The estimate from
the simulated
regression indicates that a 20 percentage point rise in eligibility under Medicaid led to a
decline in the infant mortality rate.
mortality
20
(roughly the magnitude of the expansions of the 1980s) would
on infant mortality, regardless of the measure of
eligibility
the simulated
coefficient suggests that a
of an effect of these expansions on the incidence of low birthweight, but the effect
On
Thus, the model
In the final
two colunuis, we regress
on the incidence of low birthweight and on our
on mortality
is
eligibility
the mcidence of infant
measure.
The strong
effect of
apparent even after conditioning on the positive correlation between
changes and birthweight improvements.
Thus, there
is
evidence that
eligibility for health
13
insurance improves health, as measured by
birth
outcomes.
terms of their stated goal of reducing infant mortality, the expansions of the
In
1980s were a success.'^
We
examine the heterogeneous
will explore the cost of this success below.
effects of the different types
of
eligibility
First,
however, we
policies pursued over this
period.
c) Differential
Effeas of Targeted and Broad Expansions
The bottom panel of Table
mortality rate in each state and year
3 regresses the incidence of low birthweight and the infant
on estimates of
the percentage of
under the targeted and broad expansions constructed from the CPS.
over the
full set
of years (1979-1990), while the broad regression
years that the broad expansions were in place.
women
The
is
eligible for
Medicaid
targeted regression
is
run
run over only 1987-1990, the
All the regressions include a full set of year and state
dummies.
Table 3 shows that the targeted expansions have much stronger effects on both measures of
infant health than the
broad expansions. This result
eligible, despite the fact that the estimates
was associated with a highly
strongest
on percent
regressions. Using the simulated fraaion eligible,
eligibility
is
significant
we
4.5%
when we use
the simulated fraction
eligible are equally precise across the
find that a
two
20 percentage point increase
in
decline in the incidence of low birthweight
under the targeted expansions, but with only an insignificant 0.4% decline under the broad
expansions.
Similarly, a
20 percentage point
eligibility rise
was
associated with a
7.3%
decline in
'^Our finding of stronger effects on infant mortality than on the incidence of low birthweight is
who examined the introduction of National Health Insurance
consistent with that of Hanratty (1992),
Canada, and found significant effects on mortality but mixed effects on birthweight. It does
contrast widi the fmding of Fischer (1992), who also studied the effects of the Medicaid expansions
from 1984 onwards. He found strong effects on the incidence of low birthweight for blacks, but no
in
effects
on mortality
for either race.
14
infant mortality
under the targeted expansions, but with only a 0.6% decline under the broad
The
expansions.
findings for infant mortality persist
when we
condition on the incidence of low
birthweight.'^
IV. Explaining the Heterogeneous Effects:
The
is
birth
outcome
results presented
The Role
of Medicaid
above provide a mixed message.
On
strong evidence that insurance eligibility can improve birth outcomes.
nature of the policy appears to be key.
Why
is it
that the targeted
successful than the broad expansions in improving birth outcomes?
lie in
a
number of researchers have emphasized,
programs does not automatically
translate into coverage.
only about two-thirds of women eligible for
fmd
a similar takeup rate for
effea on
fetal health
We
suggest that the answer
unemployment insurance.
individuals
if
by pregnant women.
eligibility for social
insurance and welfare
For example. Blank and Ruggles fmd
that
Eligibility
can not be expected
to
have an
they were covered by Medicaid at any point in the
under the targeted expansions
Unemployed Parent program, and those
eligible
only the "other targeted expansions" had a
birthweight.
may
take up their benefits, and Blank and Card (1991)
also attempted finer divisions of the data.
fraction eligible
the other hand, the
unless individuals takeup their newly available Medicaid benefits.
The March CPS asks
"We have
AFDC
the one hand, there
expansions were so much more
the differential takeup of these different types of eligibility changes
As
On
Takeup
For example, we have tried separating the
under AFDC or the AFDC-
into those eligible
under any other targeted expansion.
statistically significant effect
We
fmd
that
on the incidence of low
the standard errors are large, and in the estimates using the simulated
cannot rejert the hypothesis that both types of targeted expansions had similar
However
fractions eligible,
we
For infant mortality we find using simulated eligibility that both types of targeted expansions
had quite similar effects, although once again the standard errors are large.
effects.
15
1
previous year.'*
We
women made
every 100
Of women
coverage?
period, so that about
By
the year.'^
1
eligible for
15 to
old,
.5% of women
women. This
6.5% had
in the relevant
is
how many
in the entire
We
who became
an undercount of those
is,
women
for
report
some
at
point during
population would represent
full
only a lower bound, however, since some of the Medicaid
changes (those associated with the adoption of the
is
additional
age range were pregnant
takeup rate of 0.115
figure
of these expansions; that
a child in any given year during our sample
AFDC-UP
program,
Hence, the number of
covered not only pregnancy but also other conditions.
pregnant
elasticity
coverage of pregnancy,
44 years
this calculation, a
takeup by pregnant
eligibility
can therefore estimate the "takeup"
eligible for
coverage under
example)
for
women who became
changes.
all eligibility
examine the relationship between Medicaid coverage and state/year average
eligibility
using linear probability models that control for other observable characteristics, including race,
marital status,
employment
status,
and income.
CXir data set consists of
All regressions include a full set of state and year dummies.''
a 12 year period.
In the first
column of Table
4,
we
include an indicator equal to
We
under either the targeted or the broad expansions.
'*
455,774 observations over
Unfortunately, prior to
March 1988,
find that there
is
1
if
a
woman was
an increase
health insurance coverage in the
eligible
in the probability
CPS was
assigned
according to whether one received coverage under the policy held by the head of the household.
Thus, those dependents deriving coverage from outside the household were counted as uninsured.
After
March 1988, each family member was asked about
health insurance coverage from any source.
its largest effect on children below the age of 15, so that it should not
our results; furthermore, the inclusion of year dummies will capture overall changes
This questionnaire change had
significantly bias
in the
'^
nature of responses.
All
women who
In addition,
give birth
in
a year, must have been pregnant
between 2/3 and 3/4 of
the next year.
'*We use a
women whose
Hence, the percent pregnant
linear probability
model
in
in
and for computational ease with our large sample
is at
least (1
facilitate the
size.
16
some time during
pregnancies begin in one
any year
order to
at
+
that year.
year will give birth in
.67)*6.5, or 11.5%.
use of instrumental variables below,
10%
of coverage of 2.4% for every
in
rise
This estimate appears relatively high
eligibility.
cximpared to the calculations given above. The other covariates have the expected effects; Medicaid
coverage
is
less likely if individuals are white, married,
working, or high income, and the probability
of coverage increases with the number of children.
In the
we
second and third columns,
disaggregate the eligibility measure into eligibility for
a targeted expansion (over the 1979-1990 period) and eligibility for a broad expansion (over the
1987-1990 period).
We
10%
the broad expansions; a
by 2.7%, while a
10%
find that the targeted expansions
rise in the eligibility
had much larger effects on coverage than
under the targeted expansions increased coverage
under the broad expansions
rise in eligibility
is
actually estimated to decrease
coverage.
Why
did the targeted expansions have a bigger effea
the results are subjea to omitted variables bias:
there
on coverage? One
may be
AFDC
more
are both
likely to
is
that
omitted characteristics that are
correlated both with coverage and with eligibility. Suppose for example, that
were on
possibility
women whose
be eligible (because they are poor), and more
mother's
likely to
be
covered (because they are familiar with the system). The bias associated with oniitting the family's
history of welfare dependency
is
likely to
be nwre important
in the case
of the targeted expansions,
because they applied to a lower income portion of the population.
In order to address this issue,
of Medicaid coverage
in
we
each
state
in
and year;
columns
in
(4)
columns
use the simulated measure.''
we
present instrumental variables estimates of the probability
through (7) of Table
(4)
and
(S)
we
dummies
burdensome.
(it
is
The instrument
In both cases, the targeted expansions
is
the fraaion eligible
have much larger
effects
similar to instrunoenting using the full set of
equally consistent but less efficient), but
Actual eligibility remains potentially problematic
17
is
use the actual value, and in columns (6) and (7)
''Using actual eligibility in each state and year
state*year
4.
is
much
in this
less computationally
context, since state/year
than the broad expansions; in neither case
When we
1
%
is
the coefficient for the broad expansions significant.
we
use our preferred simulated instruments,
for every
10%
rise in targeted eligibility.
This elasticity appears more reasonable
calculations discussed above, and suggests that takeup
for the targeted expansions, but
We
this
was
high
fairly
low among the population
among
eligible for the
in light
the population eligible
broad expansions."
broad expansions was
First, the population eligible for the
group had higher incomes and better access
of need, the broader expansions
Second, given a
level
difficult to bring
women who
(1987) report that
for Medicaid
many low-income
even
if
to
needy; as Table 2
less
employer provided health insurance.
may have been
less effeaive.
have never received any sort of social assistance
program, either because they do not know about
women
AFDC
benefits.
It
unaware
into the
easier for
illustrated in
These two explanations clearly have differing welfare implications; the
first
need
it;
was
fairly
benign since the
the second suggests that
women who
women who
Medicaid
program
expansions because these
had more frequent mteraaions with the government assistance system, as
difference in takeup rates
be
that they can qualify
may have been
eligible for the targeted
may
It
Rymer and Adler
or because of stigma effects.
it
families and their physicians are
they do not receive
administrators to find and notify
likely to
of the
can think of two other possible explanations for lower takeup rates under the broad
expansions.
shows,
estimate that coverage increased by roughly
women
Table
2.
suggests that the
did not take up coverage were less
needed the program were not aware of
it.
recessions will increase both Medicaid eligibility and coverage.
'*As discussed above, the targeted expansions are at least partially comprised of AFDC eligibility
changes, which should have more lasting effects on coverage than pregnancy only expansions, while
the broad expansions consist of pregnancy coverage only.
It
seems unlikely
that this factor can
explain a difference in coefficients of the magnitude of those in Table 4; the majority of the total
growth
in eligibility
under the targeted expansions was accounted for by changes
coverage.
18
in
pregnancy-only
In an effort to differentiate between these two hypotheses,
models using the subsample of
We
Medicaid.
women
The
without sources of health insurance coverage other than
results are
estimates are similar to those for the
full
shown
in the final
two columns of Table
sample, and show that even
because they were not effeaively translated
the broad expansions did
eligibility failed to affect birth
into increases in
The
4.
among otherwise uninsured
on coverage while
the targeted expansions had significant effects
This suggests, therefore, that the broad expansions of
not.
reran the instrumental variables
present the results using the simulated instrument; the relative findings using the
actual instrument are similar.
women,
we
coverage, even
outcomes
among needy women.
V. Exploring the Cost-EfTectiveness of the Expansions
a) Medicaid
Payments per Infant Saved
While the Medicaid expansions of
birth
data
outcomes,
this
improvement came
on Medicaid expenditures
States report their
a cost to the Medicaid program.
to assess the cost effeaiveness
We
and outpatient hospital costs for
examine
all
total
(ic.
"We
are grateful to Killard
in
we
use
to the
Health Care Financing
to the class of
expenditures on physicians, clinics, and
non-disabled children and non-disabled/non-elderly
pregnant wonoen and infants).
15-44 year old female population
improving
of the expansions.
Unfortunately, these data are not available by type of service
population type
in
In this section,
These reports break down expenditures according
provider, and the category of recipient.
adults.
have been successful
payments made under the Medicaid program
Administration each year."
inpatient
at
the 1980s appear to
We
that year.
Adamache of Health
these data.
19
(ie.
childbirth) or by detailed
normalize the expenditures using the
All figures are in
1986
dollars;
we
state's
deflate
Services Research, Inc. for providing us with
expenditures on hospital inpatient and outpatient expenditures by the
Consumer
hospital services, and expenditures for physician and clinic services by the
services.
Once
of Medicaid
The
again,
we
eligibility,
Price Index for
CPI
for physician's
regress this measure of expenditures on the actual and simulated measures
and disaggregate by the type of expansion (targeted versus broad).
results are reported in
Table
5.
As would be expected,
the expansions significantly
increased Medicaid expenditures; regressions using the simulated eligibility measure indicate that an
additional eligible
woman was
associated with an increase in expenditures of $232 per year.
Expenditures on the targeted expansions were much larger, amounting to $449 per eligible woman.
In contrast, although the coefficient
on
eligibility
under the broad expansions
is
sizeable,
it
is
not
statistically significant.
These data can be used
woman who became
percentage point rise
rough estimate of the cost-effeaiveness of the targeted
Using the simulated measure,
Medicaid expansions.
by $449 per
to provide a
eligible
births.
Taken
estimate that Medicaid spending increased
under the targeted expansions.
in (simulated) targeted eligibility
0.04 deaths per 1000
we
We
also
fmd
that a
one
decreases the incidence of infant mortality by
together, these fmdings imply that the cost of saving a
life
through the targeted expansions was $1.73 million."
What does
this
imply for the efficiency of Medicaid policy?
This figure appears high
"This figure is calculated as follows. To generate 1000 births, given the average fenility rate
of 0.065 in our sample, would require 15,385 mothers. A one percentage point increase in eligibility
in this sample would therefore cost $69,079 ($449 for each of 153.85 women). This would reduce
the number of infant deaths by 0.04. So, to reduce the number of infant deaths by one would cost
$1 .73 million. Note that to the extent that the newly eligible women (under either type of expansion)
were getting treated for free when they were uninsured, the net cost to society of the Medicaid
expansions is lower than the costs to the Medicaid program. Saywell et al. (1989) show that the
average cost of uncompensated care for pregnancy and childbirth in 1986 in Indiana was $2668.
Subtracting this from the cost per birth of the targeted expansions lowers the cost to society per
saved to $1.66 million.
20
life
compared
to the cost
of saving a low birthweight infant through technological intervention; as
discussed above, the Office of Technological Assessment found that the average cost of saving an
infant with birthweight less than
1000 grams was $135,000.^' However,
this figure understates the
effeaive cost of high-tech interventions, for two reasons. First, over
30%
we
comparing the
only have information on expenditures per infant that lived.
In
of these infants died, and
OTA
figure to
ours, the cost per lived saved should be inflated to account for babies that used resources, but were
not saved.
Second,
described in Part
I,
serious handicaps.
we have no
there
is
Hence,
infant mortality through
information on the quality of
life
a distinct possibility that surviving
to the extent that the
improvements
of infants
who
low birthweight
infants will suffer
Medicaid expansions were effective
in fetal health, their costs (per life
As
are kept aJive.
in
reducing
of a given level of health)
will be overstated relative to high-tech interventions.
Moreover, $1.73 million
a
life.
Manning
et al.
is at
the low end of the recent range of estimates for the value of
(1989) use data from studies of willingness to pay for a small change
probability of survival to estimate that the value of a
life is
$1 .66 million.
in the
Cropper and Oates (1992)
report that studies based on compensating differentials for risk of death on the job yield a value of
life
of between $1.6 and $9.0 million; Viscusi (1992) presents a broader range, with a preferred
estimate of $12.1 million.^
have spent,
at nx)st, as
^'This and
all
much
Judged by
to
this yardstick, the targeted
Medicaid expansions appear
keep an infant alive as that person would have been willing
subsequent figures are expressed
in
1986
dollars.
Where
deflation
is
to
to
pay
necessary,
expenditure figures.
it is done using the Medical care component of the CPI, for comparison to our
The correa deflator for this exercise is not obvious, but using different deflators would not
substantially alter our conclusions.
^
Furthermore, compensating wage differentials understate the amount than an average person
would pay to reduce the risk of death, because the least risk-averse persons tend to take the riskiest
jobs.
21
^
to
be kept
alive.
Alternatively,
we can compare
the cost of saving a life through the Medicaid expansions to
the cost of keeping individuals alive through other public policy interventions.
Graham and Vaupel
(1980) report the cost of saving a life-year through 57 different public policy interventions, ranging
from mandatory
(the
seat belts or air bags to reduced emissions.
Using a
U.S. average for 1986), 38 of their 57 interventions cost
However,
the other 19 interventions cost
$30 million.^ Cropper
per consumer
producer
life
government
life
et al.
much more, and
in
life
less than
some
expectancy of 74.8 years
$1.7 million per
cases, the costs
were
life
saved.
as high as
(1992) estimate that regulations on the use of pesticides cost $60,000
saved, but that regulations on the production of pesticides cost $35 million per
saved.
Thus, while the targeted Medicaid expansions were not the most cost effective
life-saving
intervention, they
were much more
efficient than
some of
the policies
currently being pursued.
b) Evidence from Individual-Level
The
effective
results
above suggest
means of saving
relative to the
Data
that the targeted
Medicaid expansions may have been a cost
Nevertheless, the cost of doing so appears to be quite high
infant lives.
promised cost savings from prenatal care discussed
explanations for this high cost estimate.
The
first is that
women
"Relative to these other studies, our findings are subjert to
our estimates are likely to be too low, since
studies are valuing saving the lives of
we
in Part
failed to
at least
are valuing saving a
two
I.
There are two possible
make use of
(offsetting) biases.
life at birth,
workers or survey respondents.
cost-effective
First,
whereas the other
Second, our estimates may
be too high, since the infants kept alive by medical interventions may be of worse health
later in life,
as described above.
"Furthermore,
mandatory
air
do not consider the cost of the intervention to the consumer;
the government, but do increase the societal cost of saving a life.
their calculations
bags have no cost to
22
prenaial care
come
care
when
they were covered by Medicaid; thai
the infant lives that
is,
primarily through expensive hospital interventions." The second
targeted.
These alternative views obviously have very different implication
Evidence from case studies of Medicaid expansions suggest
for the high cost of saving an infant
expansion of
eligibility in
Tennessee
to
would not allow
we
Longitudinal Survey of Youth
(NLSY)
We
to
"Women may
Medicaid
The
is
die explanation
increased Medicaid enrollments, but that
Enrollment
failed to take
effects of eligibility
first is
the first trimester of pregnancy.
have
eligibility,
former
such a
at
late
do so using individual-level data from the National
examine the
expansions on two indices of prenatal care use.
doaor during
for future policy design.
(1990) found that a 1985
et al.
days prior to delivery.
thirty
not appropriately
is
attempt to distinguish between these views by examining the effect of the
expansions on the use of prenatal care.^
a
that the
it
for the effeaive use of prenatal care.
In this seaion,
saw
women
married
most of the increase took place within the
date
For example. Piper
life.
praaice, prenatal
that, in
is
not as cost effeaive as clinical studies suggest, perhaps because
is
were saved may have
under the targeted
an indicator equal to one
Early initiation of prenatal care
if
the
is
woman
important
advantage of prenatal care because they were unaware of
or because available providers were unwilling to accept Medicaid coverage.
Neither of these problems would deter the use of expensive hospital care because hospitals that
participate in
Medicare are prohibited from refusing
already in labor
(OTA, 1987b). Furthermore,
uncompensated care, accounting
for
make every
effort to enroll eligible uninsured
(Gold
aL, 1993).
Once
et
is
enrolled,
al.
women may
women
receive
in
al.
services than their
(1990) find that uninsured patients
receive less intensive hospital treatment than insured patients along a
much
et
Medicaid for the costs of childbirth
much more expensive
(1991) and Wenneker et
''The expenditure data do not offer
women
component of hospital
al., 1989). As a result,
the largest single
17.4% of these expenditures (Saywell
hospitals
uninsured counterparts; Hadley et
or from transferring, any
to treat,
childbirth
number of margins.
Both physician and hospital
insight into this question.
draw further
inferences from the relative changes in these categories, since the distinction between them is not
clear; for instance, the physician category will include the costs of deliveries done by outside
doctors, but perhaps not the costs of deliveries done by resident interns.
expenditures rose significantly under the targeted expansions.
23
It
is
difficult to
both for conducting
screenings, and for establishing a base-line for the monitoring of
initial
development (OTA, 1987b).
The second measure
sonogram during
months of the pregnancy.
ultrasound
itself
the first six
has any effea on the fetus,
of delivery as a sign that a physician
The
Since 1983,
NLSY
women
birth, including
began
in
is
we view
an indicator equal to one
While there
is
if
no evidence
woman
had a
that the use of
the receipt of an ultrasound before the point
monitoring the pregnancy.
is
1979 with a sample of 6,283
women between
the ages of 14 and 21.
have been asked bi-annual questions about the prenatal care
when
the
fetal
that
preceded each
they initiated prenatal care and questions about sonogram use.
Retrospective
information has also been collected for births before 1983.
NLSY
The
also contains
enough
information about income, family structure, and state of residence to allow us to determine whether
the
woman was
eligible for
Medicaid coverage
using the same simulation program
we
applied to the
After the exclusion of missing values,
occurred between 1979 and 1990."
sample of U.S. mothers
poor were over-sampled.
bom
to
78%
are
We
do so
data.
left
with 4759 observations on births that
NLSY
is
not a representative
age range because African- Americans, Hispanics, and the
Almost half of the
of the African-American infants,
we
CPS
important to note that the
It is
in the relevant
of the pregnancy.
in the first trimester
infants are
African-American or Hispanic, and
of the Hispanic infants, and
32%
73%
of the other infants were
mothers from the supplemental "poverty" sample.
The data
indicate that throughout
coverage of their pregnancies are poorer,
most of our sample period,
less educated,
and more
women
likely to
eligible for
be African-American or
Hispanic relative to the sample as a whole, and even relative to the subsample of
" One
shortcoming of these data
and birth outcomes took place
is
Medicaid
women
with
that the last available survey that asked about prenatal care
in the spring
of 1990.
must await the release of the 1992 survey.
24
Information about births
in the rest
of 1990
incomes below the poverty line." They also have much lower scores on the Armed Forces
QuaJiflcation
Test (AFQT),
Medicaid-eligible
eligible
women
women
are
a
standardized test of ability.^
32% more
sonogram
in the first
These comparisons highlight the
eligibility
not surprising that
is
it
likely to delay obtaining prenatal care:
20%
delay care compared to
eligible received a
Hence,
in the entire
sample.
6 months compared to
possibility that estimates
Similarly,
48%
27%
of Medicaid-
43%
of Medicaid-
of non-eligible women.
of the effects of individual Medicaid
on the usage of prenatal care may be biased towards zero by omitted variables correlated
with both eligibility and propensity to seek care.
problem, we instrument
In order to address this
individual eligibility using first both the actual and the simulated fraaion eligible measures
CPS,
calculated for this cohort of
young women.'"
expansions because the post-1987 sample size
expansions affected Medicaid coverage,
^*
it
is
We
examine only the impaa of
Also,
quite small.
seems reasonable
to
assume
from the
the targeted
since only the targeted
that the
broad expansions had
random measurement error and minimize the amount of
missing data, we use the average income in the two years preceding the birth as our measure of
income. If the woman was living with her parents then we use the parents' income less the need
standard for a family of that size (following the procedure use by the AFDC program to impute
family resources to minors living at home). Otherwise, we use the sum of the woman's own income,
the spouse or partner's income, and "other" income.
The use of this measure also avoids the
In order to attenuate the effects of
imputation of eligibility for prenatal care on the basis of temporary income losses suffered after the
birth.
"
Since the
AFQT
was administered
normalized using the mother's age.
to all the
Some
readers
measure of background and education, rather than
''^e problem
that individual eligibility
outcomes has been noted
in the
may
women
may
as a
at the
same point
prefer to regard the
measure of native
wider social insurance
literature.
were
summary
ability.
For example, unemployment
may confound
of those benefits on unemployment durations (Meyer, 1989).
interpretation of the effect
In that context,
control for the wage, or nonlinear functions of the wage, directly.
it
may be
Since Medicaid
complicated function of a large number of individual characteristics,
25
as a
be correlated with unobserved determinants of
insurance benefits are a fiinaion of lagged wages, and this
variables strategy.
in time, scores
AFQT
we
easiest to
eligibility is a
prefer the instrumental
little
impact on the use of prenatal care.
We
estimate logits models which include exogenous characteristics of the mother and child,
in addition to a full set
dummies.
As discussed above,
characteristics of states that
state fixed effects
state (California),
We
of year dummies."
demand
also estimate
models with and without a
set
of
state
the inclusion of state fixed effects controls for time invariant
may be
correlated with state Medicaid policy.
a lot of our data
9 states represented
in
--
However, models with
although there are over 600 observations for the largest
our sample have fewer than 15 observations.
We
do not
include fixed effects for these 9 states, so together they form the omitted "state".
The instrumental
effects,
we
find
variables results are presented in Table 6.
some evidence
In the
models without
state
expansions led to increased use of early prenatal
that the targeted
Using the simulated instrument, our estimates imply
that eligibility for
Medicaid under the
targeted expansions decreased the probability that the initiation of prenatal care
was delayed beyond
care.
the first trimester by
first six
13%. There
months; the coefficient
is
is
no
statistically significant effect
wrong-signed. Once
on
the use of
sonograms
state effects are included in the
in the
model, none
of the results are significant, although the sonogram result for the simulated instrument becomes
right-signed (eligibility raises the probability of a
state effects
themselves are not jointly
the targeted expansions
were successful
sonogram
in the first six
statistically significant.
in increasing early contacts
months).
However,
the
Thus, the evidence on whether
with physicians
is
mixed. These
data are unable to definitively resolve whether prenatal care did or did not rise under the targeted
expansions.
VI. DiscussioD
"We use the predicted
and Conclusions
value of eligibility from the
first
stage linear probability model (individual
regressed on state/year eligibility) directly in the second stage logit. This procedure
consistent but not efficient, and our standard errors are somewhat under-stated as a result.
eligibility
26
is
A
key question for health reform
improvements
in health.
While
healthy,
whether covering the uninsured will actually lead
number of
a
studies have
surmount the problem
health, they are generally unable to
less
is
in the
Medicaid
eligibility
of pregnant
that the uninsured are in
that the uninsured
Our approach
independent of insurance status."
exogenous changes
shown
women
to
in the
worse
may be fundamentally
problem
this
to
1980s.
to
is
examine
Judged by the most
frequently used indicator of birth outcomes, the Medicaid expansions were a great success; infant
mortality
women
was reduced by
7%
for every
eligible for Medicaid.
relative to the value
of a
life
limited, along
low income
at the effects
little
clinical sojdies
of
women
life
to obtain such care, or
studies suggest.
A
have been reasonable,
broad expansions of Medicaid
We
lives.
eligibility to ail
among needy
were unable
to resolve
whether prenatal care
fruitful direction for future research
is
women.
(otherwise uninsured)
under the targeted expansions was much higher than
of prenatal care.
44 year old
effea, primarily because they were not effectively
translated into increased Medicaid coverage, even
by
to
to
of the Medicaid expansions suggests that their success
First, the later
appear to have had
Second, the cost of saving a
fraaion of 15
Moreover, the cost of the policy seems
two dimensions.
women
rise in the
or the cost of other government policies designed to save
However, a closer look
was
20 percentage point
whether
this
that suggested
was due
to the failure
simply not as cost effective as clinical
would be
to investigate
more
carefully,
perhaps through additional case studies of targeted expansions, the utilization of different types of
prenatal and hospital care in response to the increased availability of public insurance.
Thus, our research offers two insights for the design of insurance policy.
First, increased
insurance eligibility can improve health outcomes, and the cost of doing so appears to be lower than
'^ne
study which does surmount this problem
is
that of Lurie et al. (1984),
which finds
that
individuals (exogenously) terminated from Medicaid due to California budget cutbacks saw reductions
in their
self-reported health.
27
many
estimates of the value of a
effective
is
to ensure that the
qualified for
government
newly
Second, the crucial
is
moving
making insurance policy
Groups who have not previously
programs may be particularly hard
to
to reach.
Fortunately,
address this problem (National Governor's Association,
Several states have adopted public relations campaigns with themes such as
Baby" (Utah) or "Baby Love" (North Carolina)
In fact, in contrast to the findings
et al.
step in
initial
eligible takeup their benefits.
social insurance
public policy at the state level
1992).
life.
accompany expansions
to
from case study of Tennessee
that
in
"Baby Your
Medicaid
eligibility.
were discussed above, Buescher
(1991) find that the North Carolina program had significant positive effects on the utilization
of prenatal care and on birth outcomes.
To
the extent that informational problems are to blame,
national health insurance of the type proposed by President Clinton
takeup problem since there would be no question of
would presumably surmount
this
eligibility.
Another question which our research does not resolve
effective in improving the health of other populations.
is
whether insurance policy
Birth outcomes
may
will be
be a particularly
responsive margin to health interventions, relative to other medical problems later in
life.
It
would
be useful to extend the methodology developed here to explore the effect of exogenous changes
insurance coverage on health at other points in the
28
life
cycle.
in
Table
ST
1: Eligibility
by State over Time
1979
1986
Any Exp.
Any Exp.
Any Exp.
Targeted
Broad
AL
.146
.214
.342
.148
.194
AK
.058
.187
.243
.216
.026
AR
.123
.151
.307
.137
.169
CA
.189
.252
.423
.271
.151
CO
.095
.192
.236
107
.128
CT
.154
.165
.253
.161
.091
DE
.114
.114
.243
.092
.150
DC
.286
.215
.454
.308
.145
FL
.094
.133
.361
.170
.190
GA
.090
.178
.280
.200
.079
HI
.202
.179
.389
.522
.167
ID
.095
.168
.300
.159
.141
IL
.171
.186
.268
.181
.086
IN
.081
.137
.300
.140
.159
lA
.103
.165
.380
.136
.243
KS
.117
.157
.293
.131
.161
KY
.119
.154
.396
.210
.186
LA
.164
.225
.391
.77.6
.164
ME
.141
.200
.420
.222
.198
MD
.137
.167
.350
.156
.194
MA
.156
.165
.352
.209
.142
MI
.173
.225
.390
.228
.161
MN
.099
.209
.375
.184
.191
MS
.111
.233
.567
.233
.334
MO
.092
.155
.273
.109
.164
MT
.165
.137
.334
.177
.157
NfE
.133
.153
.233
.114
.118
NV
.095
.088
.254
.123
.131
NH
.096
.072
.184
.128
.055
1990
29
ST
1979
1986
Any Exp.
Any Exp.
Any Exp.
Targeted
Broad
NJ
.157
.149
.232
.157
.074
NM
.110
.147
.347
.137
.210
NY
.238
.242
.411
.257
.153
NC
.106
.150
.364
.202
.162
ND
.110
.134
.279
.146
.133
OH
.131
.187
.278
.172
.105
OK
.139
.162
.314
.198
.116
OR
.161
.155
.264
.145
.119
PA
.138
.164
.293
.182
.111
RI
.197
.152
.352
.187
.164
SC
.150
.228
.401
.160
.241
SD
.110
.149
.277
.131
.146
TN
.104
.162
.361
.216
.145
TX
.079
.143
.318
.141
.176
UT
.132
.255
.243
.140
.102
VT
.228
.234
.396
.239
.157
VA
.129
.245
.271
.136
.135
.170
.242
.262
.165
.097
.105
.215
.408
.206
.201
WI
.126
.232
.271
.175
.096
WY
.068
.140
.245
.127
.118
WA
WV
Notes: Figures are fractioo of 15-44 year old
(columns
1,2,
and
3), targeted
1990
womeo
ID
each state/year
eligil)le for
Medicaid under
expansions only (column 4), and broad expansions only (column
March 1980, 1987, and 1991 CPS.
30
5).
ai
Tabulated from
Table
Population Covered Under
"Targeted" and "Broad" Expansions, CPS Data
2: Characteristics of the
Characteristic
Sample
Targeted
Expansions
Expansions
38,403
18,775
22,164
(29,768)
(27,676)
(22,232)
24.9%
99.1%
38.5%
Full
Family Income
Poor
Number of Kids
Broad
1.26
1.34
1.31
(1.26)
(1.41)
(1.39)
839.%
71.7%
80.9%
29.9
25.8
28.4
(8.35)
(7.54)
(7.67)
Married
51.8%
12.7%
38.0%
Working
74.7%
52.6%
72.4%
6.1%
26.2%
5.08%
Uninsured
16.4%
30.1%
31.7%
Employer-Provided HI
33.2%
6.91%
25.4%
Medicaid
10.1%
37.6%
12.8%
White
Age
Received Public
Assistance
N9t?§: Data from 1990
previous year; poor
is
saimple
of CPS.
Working
is
defined as working at least
family income below poverty line for that family size.
31
Table
3:
OLS
Regressions of
Low
Birthweight and Infant Mortality on Eligibility
Using VUai Suidstics
Low
DaU
Each State and Year
Tor
Birthweight
Infant Mortality
Actual
Simul.
Actual
Simul.
Actual
Simul
Elig.
Elig
Elig
Elig.
Elig.
Elig.
Models Using Fraction Eligible Under Any Expansion
Fraction
^.211
-7.431
-2.093
-3.458
-1.926
-3.171
Eligible
(2.886)
(3.848)
(0.819)
(1.090)
(0.813)
(1.085)
Low
Birthweight
Intercept
Adjusted R'
0.040
0.039
(0.012)
(0.012)
74.29
75.25
9.162
9.571
6.206
6.662
(1.145)
(1.371
(0.325)
(0.388)
(0.957)
(0.990)
.969
.969
.906
.906
.907
.908
Models Using Fraction Eligible Under Targeted Expansions
Fraction
-7.749
-15.37
-1.752
-3.989
-1.439
-3.391
Ehgible
(4.123)
(5.920)
(1.176)
(1.689)
(1.170)
(1.685)
Low
Birthweight
Intercept
Adjusted R^
0.040
0.039
(0.012)
(0.012)
74.13
75.49
8.814
9.201
5.821
6.267
(0.983)
(1.226)
(0.280)
(0.350)
(0.945)
(0.984)
.969
.969
.905
.906
.907
.907
Models Using Fraction Eligible Under Broad Expansions
Fraction
Eligible
Low
Adjusted R'
:
-1.451
-0.108
-0.336
-0.067
-0.232
(2.458)
(6.178)
(0.699)
(1.757)
(0.679)
(1.706)
Birthweight
Intercept
Notes
-0.565
All regressions
0.072
0.072
(0.023)
(0.023)
74.35
74.35
8.440
8.447
3.081
3.089
(1.406)
(1.397)
(0.400)
(0.397)
(1.743)
(1.742)
.981
.981
.910
.yi\j
.910
.»iu
.916
mcluded
state
expansion and targeted expansions;
and
N =
y«
iS^
Standard errors in parenth(leses.
2(X) for broad expansions.
32
.916
N = 600
for
any
a.
^-K
£^1
O O
o d
C/5
^^
^-S
O O
d d
O
9 d
<^
00
o
9
2,
00
- O
d d
« 2:1
i/3
o
d
O
"5 5^
o
90
9 o
o d
O
9 o
"1
</^
a
3
?
o
£
5"
o
9d 90 90 9o
o
<::>
d^
9o
^
^
U
fa*
u
s
o
^1
>
2,
3
— O
00
S3
S
li
od
d d
CD
d
2—
gSS
92,
"1
f^
—
o
90
—
^<N ""M —
§g 2g z§ 2g 2
90 90 90 dd 9d
£!fN
50
W-1
?
3
-3
o
u
o
9 O
S
9 d
9S
5i
u
e
u
w
m
O <N
O
£ >3
002-
8S So 98 2g =8 38
d d ^2°2'
*'8-
^S ^S ^S
'2- 5
—
s.
IT)
O
t
.^1
1
ro
(N
u
—
u
.SF
—
S8
O
d d
s
a
2
o
p>i
d d
2-9
2-
<N
S? fn
98 S8 98
90 9 d 9 d
vr
—
in
r—
t-
9
d d
9
o o
d d
° 8
9 d
''^
iO
.
•Q ii
•3
&
3 E
"Sc^
^=3
n
f^
of-
98 00
9d dd
2
>-^
.,^^
i*
H - !2 <^ ^ <^
•? o
98 28 = 8
9292- ''S- 9 d
>»^
-..^
^.ir
3
-3
:$
2
II
i
u
^?
•a
3 ^
s
9v
QO
t<
d d
2
o
_
58 ^§ 88 o i Sid
9d 9d 9d 9
2,
J58
<^
8
o d
d
= § s
9 d
_ V — *
—
o_
S8
S8
-8
S8 -8
98
08
do 9d 9d 9d ? 9- 2- 9 2
If—
31
d d
"^S"
fs
->
p^i
3
—
>
a
V
s
IT)
p~
I
i
g 3
o
1
^E
5 .2?
|.S
|S
<
b-
ca
2*
a-
._
^
<
V
<
O
u
1
!
a
ao
E
a
^
2
^
u.
2 *
« ••
u >
3
S
3
z
Z
?
Table
5:
Medicaid Eligibility and Medicaid Payments:
Evidence from HCFA Data
Measure of Fraction
Elig:
Overall Eligibility
Targeted Expansions
Broad Expansions
Actual
Simulated
.206
.232
(.077)
(.102)
.357
.449
(.110)
(.156)
.137
.262
(.154)
(.187)
Notes This table shows the coefficient on various Medicaid eligibility variables from regressions
including year and state dummies. For example, the coefficient under "actual" in the targeted
expansions row refers to the coefficient on the actual fraction made eligible under the targeted
expansions. Standard errors in parentheses. Dependent variable is payments in thousands of
:
1986 dollars per 15-44 year old woman.
34
\
i
i
"3
O
r4
so r^
r^
CM
—
—
O
O
— o
<N
—
O
O
— O
vn
O
in
u
3
•J
<
p~
On
SO
rj in
\0 in
Pv|
r^t
r-1
r-i
CM
— O
O
z
V
1
3
3
B«
^
0\ (N
ro 00
O O
r-
O
—
m —
—
o o
•*
o
1^ —
u-i
<N ro
00 CN
oO
m
o 00
o
irl
>
—
o o
NO
fN 00
—
o
00 —
r-1
vO
so
m
00
—
m
vo
<t
—
*""
SO
—
<N 00
rM
CTv
f<)
^
5
OS
00 eN
OO
—
•*
(^ 00
— o
On
cm OS
—
r-i
a
s
•s
«
§
s
*g
3
ZV
e
e
_2
3
E
m
>n
CM
On
<N <N
'
SO 00
ss
^*^
^"^
^"^
f^
ss
1
'^-^
i
a
C
,^^
C^
>.
"sS
r~ •—
2
«N
/—
^nm^
\0 00
S^
00
ro <N
1/^
-^
<
^
8
t«
1
E
3
—
,^_^
o
oo
in
00 •*
so
OS
vn 00
m
1
NO r^
00
Q
—
.^M^
r^
S^
r-
:::S
—
*
yrt
O
^
oo
00
—*
SO •*
On iO
00
is
5
2S
»n •la
OS •^
"^
SO <s
00 <N
oo
m
— 00
o
NO (N
00 <N
<N
SO 00
m
— —
^.^
V5
o
^
I
ow
o
in
t3
O
r*>
O
oo
OO
—
—O
tr.
u
^
(X
</9
ka
e
1>
E
5
a
4J
s
C/3
fl
w
4)
E
eted
M £?
:^
W H
ca
«J
k>
u <i
?3
e
25
^
U iZ
1/1
c
"3
o
is
ii
o
19
it
s<
a.
1/1
o
&
u.
<
rO
—o
—
—
— 00
o
vO
«0 00
4>
a>
00
iri
o
c
V5
VO
e
o
>n r-
so
00
—
O
—
vO
—
r— o
o
—_ oo
o
m
00
VO
C «
2 -s
o
00
in
>/^
rvi
"O
CM fM
IT)
S
4)
2 E
.£
W) 00
so
3 o
=s
^2
O^ 30
—
— 00
2
(N
oo
3 00
o
W->
= s
l|
o
.£
00
o m
t/5
s«
g-a
Ov 00
s
— o
ON
S
— O
"8
Hi
o
00
r~ 00
2o
c
VO
4>
E
o
(?v
00
— 2
i'
o
00
—
25w
<»•
00
r— O
IT)
^
1
IT)
vO
«^
w'
'
s Se
to
r-
S
r- 00
SO
>>
VO
5
'•5
00 <o
IT) <N
00
—
r-,
oo
O O
^ 00
—
p
=
P
<N
<
r»
00
—
•^ oo
pp
946
3813
U 3
S,s
00
9> 00
•<«
c
4>
Income
e
4>
s s
^ 1
E S
o
8 Tf
R 1
>
o
a.
lember
Du Du
Age
unple
rban
Poverty
2
§^
= 2
S * I
>»
»
o
1s
«
-S
00
o
tI 8^
1
o
E
>> c
i2 •=
00
1
Q
o
;;o
3
ear
?3
s
>
ate
ca
o—
00
.-5
21
(N
4>
III
4>
E E
3 3
zz
wj
4>
<3
U
^
4)
113
^
.3
u]
1985
year
Figure
ID:
Low Birthweignt
in
the
1980s
35 9
1
u
S
J
3
i
-
1
-T
T
25
/
-
/
/
/
2
/
-
15 ^
1
^^-"^
979
Figure
1982
2a:
1985
year
1988
1990
Medicaid EligiDility Trends
c
"T
in
6
-
10
u
.5
-
4 -
1979
Figure 2a: Medicaid EligiDility Trends for <2>^Poverrv
References
Alan Guttmacher
Institute.
The Financing of Matemitv Care
in the
United States
.
New
York: AGI, 1987.
American Public Welfare Association. "Voluntary Contributions and Provider-Specific Taxes: Survey
Results". Washington, D.C.: APWA, 1991.
Buehler, J.W.,
J.
Kleinman, and C. Hogue. "Birth Weight-Specific Infant Mortality, U.S., 1960 and
102 #2, 1985, 3564-3567.
1980," Public Health Reports
Buescher, et
al.
Outcomes
.
"An Evaluation of
in
the
Impaa of
Maternity Care Coordination on Medicaid Birth
North Carolina," American Journal of Public Health
Blank, Rebecca and David Card.
"Recent Trends
of Eligibility
Is
There an
November 1991, 1157-1190.
"When Do Women Use
vs. Participation,"
81, 1991, 1625-1629.
Insured and Uninsured Unemployment:
in
Explanation?," Quarterly Journal of Economics 106,
Blank, Rebecca and Patricia Ruggles.
.
AFDC &
Food Stamps? The Dynamics
mimeo, July 1993.
Braveman, Paula, et al. "Adverse Outcomes and Lack of Health Insurance Among Newborns in an EightCounty Area of California, 1982 to 1986," The New England Journal of Medicine 321 #8, August
24, 1989.508-513.
.
Buehler, J.W.,
J.
Kleinman, C. Hogue. "Birth Weight-Specific Infant Mortality, United
States,
1960 and
1980," Public Health Reports 102 Kl, 1985, 3564-3567.
Chaikind, Stephen and Hope Corman. "The Special Education Costs of
Low
Birthweight,
"
NBER
Working
Paper #3461, October 1990.
Children's Defense Fund.
The
Deficit
Reduaion Act of 1984
.
Washington, D.C.: Children's Defense
Fund. 1984.
Creasy, R.K., B.A.
Gummer, and G.C.
Obstetrics and
Gynecology
.
Liggins, "System in Predicting Spontaneous Preterm Birth,"
6, 1980, 692ff.
Corman. Hope, Theodore Joyce, and Michael Grossman.
United States', Journal of
Human
Resources 22,
.
"Birth
Summer
Outcome Produaion
functions in the
1987, 339-360.
"Environmental Economics:
Cropper, Maureen L., and Wallace E. Oates.
675-740.
Economic Literature 30, June 1992,
A
Survey," Journal of
.
Cropper, Maureen L.
et al.
.
"The Determinant of Pesticide Regulation:
Decision-Making," Journal of
Political
Economy
.
Statistical
Nancy Cole. "Welfare and Child Health: The Link Between AFDC
Weight," The American Economic Review 283 #3, September 1993.
Currie, Janet and
Birth
A
.
37
Analysis of
EPA
100, 1992, 175-197.
Participation and
.
"Weighing the Evidence: An Empirical Analysis of the Expansions of Medicaid
Coverage for Pregnant Women, 1984-1990". Thesis, Harvard College, 1992.
Fischer, Michael.
Donna
Frank, Richard,
Storbino, David Salkever, Catherine Jackson. "Updated Estimates of the Impact
of Prenatal Care on Birthweight Outcomes by Race,"
NBER Working
Paper #3624, February 1991
Gold, Rachel, Susheela Singh and Jennifer Frost. "The Medicaid Eligibility Expansions for Pregnant
Women: Evaluating the Strength of State Implementation Efforts", Familv Planning Perspeaives
25 M5, September/October 1993, 196-207.
.
Government Accounting
Medicaid Expansions: Coverage Improves but State Fiscal Problems
Washington, D.C.: GAO, June 1991.
Office.
Jeopardize Continued Progress
.
Graham, John D., and James W. Vaupel.
Analvsis
1
.
"Value of a Life: What Difference Does
Make?," Risk
it
#1, 1981,89-95.
Grossman, Michael, and Steven Jacobowitz. "Variations In Infant Mortality Rates Among Counties of the
United States: The Roles of Public Policies and Programs," Demography 18, November 1981,
.
695-713.
Grossman, Michael and Theodore Joyce. "Unobservables, Pregnancy Resolutions and Birth Weight
Production Functions in New York City," Journal of Political Economy 98 #5, October 1990, 983.
1007.
Gruber, Jonathan.
"The Effect of Price Shopping
in
Medical Markets: Hospital Responses
"Canadian National Health Insurance and Infant Health".
Hanratty, Maria.
to
PPOs
in
Economics
California," forthcoming. Journal of Health
Mimeo, Cornell
University,
1992.
Harris, Jeffrey.
of Health
"Prenatal Medical Care and Infant Mortality," in Victor Fuchs, ed..
.
Chicago: University of Chicago Press, 1982.
Health Care Financing Administration.
D.C.:
HCFA,
Analvsis of
,<
?taTt
Mg^i?"d Program
Characteristics
Hill, Ian.
fft at
Hill, Ian.
.
.
Washington,
1982, 1983, 1984.
Health Insurance Association of America. The Cost of Maternity Care and Childb irth
1989
Economic Aspects
in the
United States.
Washington, D.C.: HIAA, 1990.
Medicaid
Eligibilitv:
? P?';P"Pfft1^
A
Descriptive Report of
OBRA. TEFRA.
and
DEFRA
Provisions, and
Washington, D.C.: National Governors Association, 1985.
Broadening Medicaid Coverage of Pregnant
Women
and Children: State Policy Responses.
Washington, D.C.: National Governors Association, 1987.
Horbar, Jeffrey, Elizabeth Wright, Lynn Onstad
et al.,
38
"Decreasing Mortality Associated with the
Introduction of Surfactant Therapy:
Grams
Hutchens,
at Birth," Pediatrics .
tfl,
An
Observational Study of Neonates Weighing 601 to 1300
August 1993, 191-196.
"AFDC
and the
Birthweight . Washington D.C.: National
Academy
George Jakubson, and Saul Schwartz (1989).
Robert,
Subfamilies," Journal of
Institute
92
Human
of Medicine, Preventing
Low
Formation of
Resources 24, 1989, 599-628.
.
Press, 1985.
"Major Changes in State Medicaid and Indigent Care Programs ".
Washington, D.C.: The George Washington University, various years.
Intergovernmental Health Policy Projea.
Eds. Debra
J.
Lipson
et al.
Manning, Williard G. et al. "The Taxes of Sin: Do Smokers and Drinkers Pay Their Way?," Journal of
the American Medical Association 261 #11, March 17, 1989, 1604-1609.
.
McCormick, Barbara,
at
"The Health and Developmental Status of Very-Low-Birthweight Children
School Age," Journal of the American Medical Association 267, 1992.
et al.,
.
"A Quasi-Experimental Approach
Working Paper #3159, 1989.
Meyer, Bruce.
"Incentive Effects of the U.S.
Robert.
Moffitt,
Literature
.
30,
March 1992,
to die Effects of
Welfare System:
A
NBER
Review," Journal of Economic
1-61.
Piper, J.M. et al. "Effects of Medicaid Eligibility Expansion
in
Unemployment Insurance,"
on Prenatal Care and Pregnancy Outcomes
Tennessee," Journal of the American Medical Association . 264, 1990, 2219-2223.
Rosenzweig. Mark and T. Paul Schultz. "The Behavior of Mothers as Inputs to Child Health: The
Determinants of Birth Weight, Gestation, and Rate of Fetal Growth," in Economic Aspects of
Health
.
Victor Fuchs (ed). University of Chicago Press: Chicago, 1982.
Rosenzweig, Mark and T. Paul Schultz. "Estimating a Household Production Function: Heterogeneity, the
Demand
for Health Inputs,
and Their Effects on Birth Weight," Journal of
Political
Economy
.
91,
October 1983, 723-746.
Rosenzweig, Mark and T.
Replication,"
Paul
Schultz.
"The
Stability
The Journal of Human Resources
.
Produaion Technology,
of Household
A
23, Fall 1988, 535-549.
Rymer, Marilyn and Gerald Adler. "Children and Medicaid: The Experience
Financing Review 9#1, Fall 1987, 1-20.
in
Four
States," Health
Care
.
Saywell,
et
al.
"Hospital
and
Patient
Characteristics
Implications," Journal of Health Politics. Policv. and
of Uncompensated
Uw
.
Hospital
Care:
Policy
14, 1989, 287-307.
Schwartz, Rachel. "What Price Prematurity," Familv Planning Perspectives 21, July/August 1989, 170.
174.
39
U.S. Department of Health and
to Familiea
U.S.
Human
Services (various years).
With Dependent Children
.
Washington, D.C.:
Characteristics of State Plans for Aid
US DHHS.
House of Representatives, Green Book 1992: Background Mate rial on Programs Under
Jurisdiction of the Committee on Wavs and Means Washington D.C.: GPO, 1992
the
.
U.S.
Office of Technology
Assessment.
Washington, D.C.: GPO, 1987a.
Healthv
Children:
Investing
in
the
Funire
.
OTA-H-344,
U.S. Office of Technology Assessment. Neonatal Intensive Care for Low Birthweight Infants: Costs and
Effectiveness . OTA-HCS-38, Washington, D.C.: GPO, 1987b.
Viscusi, Kip.
Weitz, Judith.
FrltH l
T^? '\ ^^^}
New
York: Oxford University Press, 1992.
Testimony before House of Representatives on
DEFRA
(PL 98-369), 1983.
Yelowitz, Aaron. "The Medicaid Notch, Labor Supply and Welfare Participation: Evidence from Eligibility
Expansions,"
MIT mimeo,
July 1993.
40
Appendix:
Eligibility
Simuiatioas
During the entire period studied, 1979-90, the fundamental basis for eligibility for Medicaid was
AFDC. The core of our eligibility calculation was therefore a simulation of eligibility for
welfare receipt. In order to qualify for AFDC, a family must pass three tests: their gross income must be
below a multiple of the state's needs standard (this test was applied from 1982 onwards only);" their
gross income less certain disregards for work expenses and child care must be below the state's needs
standard; and their gross income less certain disregards less a portion of their earnings must be below the
state's payment standard.
receipt of
The exaa definition of a family unit is the first source of difficulty in making this calculation. If
minor
(which we define as less than age 19) is living with her parents, then a ponion of the parents'
a
income is deemed to that individual in making the eligibility calculation. This fraction is calculated by
subtracting from family income the needs standard for a family of that size. If the individual is age 19 or
above, then the treatment of family resources is less clear, and varies across states; see Hutchens et al.
(1989) for a description of these differing treatments. We assume, following the praaice of the majority
of the states, that the parent's resources are ignored
Before
OBRA
Before
OBRA
if
the individual
is
not a minor.
was no standardized allowance for work and child care expenses under
AFDC. Beginning in October 1981, the allowance was standardized to be $75 per month for work
expenses, and a maximum of $160 per child for child care costs. This was not changed until the Family
Support Act of 1988, which raised the allowances to $90 for work expenses and $175 per child for child
care, effective October, 1989. For years before 1982, we assume that states allowed the same level of
disregards as was designated under OBRA.
'81, there
'81
,
there
was a work
incentive feature of
AFDC known as
"30 and one-third"
.
This
$30 of earned income, and one third of die remainder, while
Since individuals must qualify for an AFDC payment
in order to receive Medicaid, this meant that the individual's income, less disregards, and less $30 and onethird, was compared to the payment standard for the third test described above. From 1982 to 1984, this
incentive was limited to the first four nwnths that individuals were on AFEKD. In order to model this, we
assume that for our sample it is their first full year on AFDC, so that they get $120 and 1/9 of their
earnings for the year. From 1985 onwards, individuals who would have lost Medicaid due to the end of
the $30 and 1/3 rule after 4 months were allowed to remain on Medicaid for an additional 9 to 15 months
(the length was at state discretion). We modelled this as amounting to a fiill 30 and 1/3 exclusion after
1984, parallel to the pre-1982 rules. Our objea was to consistently nnodel die maximum amount that an
incentive allowed families to keep the first
the other two-thirds led to reduced
AFDC
individual could have received under
F'mally, a
key
restriction
benefits.
AFDC.
on the
receipt of
AFDC
is
family structure.
In
all states,
single
women
one child are eligible. In addition, in some states, married women with an unemployed spouse
Eligibility for AFDC-UP conditions on both current
are eligible under the "AFDC-UP" program.
employment status and work history. Lacking longitudinal data on work histories, we assume that families
are eligible if the state has a program and the spouse had worked less dian 40 weeks in die previous year.
with
at least
"From 1982
to 1984, this multiple
was
1.5; ft-om
41
1985 onwards, die multiple was 1.85.
Since families eligible for
AFDC-UP
malce up only a small fraction of the overall
AFDC
population, this
should not greatly affea our estimates
For most Medicaid recipients, AFDC receipt summarizes the necessary conditions for Medicaid
However, from the inception of the program, there have been a number of special options, to
be used at state discretion, which allow further means of eligibility for certain populations, and, in
eligibility.
particular, for pregnant
The
who
first
women.
mothers who are pregnant for the first time,
Before 1982, a number of states had first-time pregnancy
programs, which covered women from some point in their pregnancy even if
of these are
state
programs
are therefore not yet eligible for
options under their
ATOC
they had no other children.
although
some
states
the first trimester.
most
women were covered ftx)m the point of medical verification,
later point; we only counted those states that covered women during
Generally,
covered them
In
to cover single
AFDC.
at
a
states, the states
counted only the nwther
in the family unit size (i.e. the size
of the family was one), although a few states counted the future child as well. The states which had these
programs are identified in U.S. Department of Health and Human Services (various years).
After OBRA '81, states with these programs were no longer allowed to cover women before the
month of pregnancy. However, states were offered the option of deeming first-time pregnant women
eligible for Medicaid from the moment of verification, even though AFDC payments had not yet begun
to flow. As with most Medicaid-only options, information about which states took up this c^nion is spotty
and often contradictory. Based on information in Health Care Fmancing Administration (1982,1983,1984),
Weitz (1983), and Hill (1985, 1987), we came up mAi a list of states which appeared to have taken up this
option in each year. We also assumed that, under this program, states did not count the unborn child in
sixth
determining the size of the assistance unit.
Under
DEFRA
which was effective from 1985 onwards, states were required to cover "a
eligible for Aid to Families with Dependent Children ... as if her child had
bom
was
living
with
been
and
her ..., [ifl such pregnancy has been medically verified' (P.L. 98-369, Sec.
2361, Subtitle B). Thus, first-time pregnant women are covered in all states frx>m 1985 onwards, and the
effective family size for AFDC receipt is increased by one for all pregnant women.
pregnant
'84,
woman who would
Another unportant
would qualify
for
AFDC
structure (ie. children in
of such children
(ie.
be
state
Medicaid option
is
coverage of 'Ribicoff children",
under income considerations alone, but
poor two-parent families). Before 1981,
only those In institutions, only those
in
who would
states could
who
are children
who
not qualify due to family
cover selected categories
two-parent families, etc.), or
all
categories,
up to age 21 From 1982 onwards, they were also offered the option of restricting die coverage of children
between 18 and 21 years of age. Thus, for teens in our sample in states with comprehensive Ribicoff
children programs, we ignored tbe family structure requirements of AFDC and only screened on
.
income.**
which took up die Ribicoff option to cover poor
children in two-parent families. However, the data on this program are spotty, and this category is
never presented. Thus, we only included those states which were listed as having programs that
The data on this program are from HCFA
covered all categories of Ribicoff children.
**In theory,
we
should do this for any
state
42
Furthermore, although OBRA '81 restricted the AFDC coverage of unborn children, states were
given the option of covering unborn children under their Ribicoff children programs. This amounted to
covering pregnant women whose income qualified them for AFDC, regardless of family structure. We
HCFA
gathered information on this Ribicoff option from
modelled
it
as being in place
(1982,1983, 1984) and from Weitz (1983), and
from 1982 onwards.
Medicaid program, states have had the option of covering
them for AFbc but who were not single mothers. This little
In addition, since the inception of the
all
women whose income
pregnant
ioiown option
is
only discussed
qualified
Weitz (1983), Children's Defense Fund (1984), and
in
Hill (1987); the last
of these sources gives the dates of the inception of these programs.
Each of the three options
restrictions for pregnant
women)
just listed (Ribicoff teens, Ribicoff unborn, and no family structure
apply only to the pregnant
women
herself, or only to her unborn child,
would be one. However, we assumed that
had more than one of these programs, the family unit size was increased by the future child.
so the family unit size (for example) for a fu"st time pregnancy
when
states
Needy program, which is designed to cover
individuals who meet the family structure requirements for AFDC and whose gross resources are above
AFDC levels, but whose high medical expenditures bring their net resources below some certain minimal
level.
States who take up this option may establish Medically Needy thresholds that are no more than
133% of the state's AFDC needs standard. Individuals can then 'spend down' to these thresholds by
subtracting their medical expenditures from their gross income; if they do, Medicaid will pay the remainder
The fmal
state
option of importance
is
the Medically
of their expenditures."
While this program may be
whose income is somewhat above
option
First,
is
we
quite useful for covering the birth expenditures of a pregnant
the
it
is
We
unclear
how
important the Medically Needy
therefore model this
program
in
two ways.
is, this
program
women whose gross incomes are between the AFDC threshold
Second, we subtraa from gross income the expected global fee for
pay for the medical costs of
and the Medically Needy threshold.
an obstetrician, itom Alan Guttmacher
pregnant mother for their global fee
qualify for the Medically
former
cutoff,
simply compare the woman's net income to the Medically Needy threshold; that
will help to
bill to
AFDC
for covering her prenatal care expenditures.
woman
Institute
(1987).^ The notion here
at the first prenatal care visit,
Needy program. Since
is
that, if obstetricians bill the
then the mother will be able to use this
the measures
were very
similar,
we
only used the
in the analysis.
Beginning with
DEFRA
'84, the Federal
government began a
series
of mandates which extended
(1982,1983,1984) and Hill (1987).
"The
model
time frame over which such spend-down occurs varies across the states, and
we do
not
it.
**The global fee
natal care.
is
the
bill
for
all
obstetric services, including pre-natal care, delivery, and post-
This fee averaged $830 in 1986, and
physician's services component of the
Consumer
is
deflated (and inflated) to other years using the
Price Index.
43
DEFRA '84 txKiuded two features: mandatory coverage of
above),
and mandatory coverage of pregnant women in AFDCwomen under AFbC (described
type families, even if the state did not have an AFDC-UP program. COBRA '85 then mandated that
the Medicaid
covenfe of prefnant women.
pregnant
UP
women who met the AFDC
resource standards were eligible regardless of family structure (similar
This law was effective in July, 1986. Since we are focusing
to the state programs described above).
pregnant
primarily on the effect of Medicaid on early prenatal care,
we model
this
law as being effective from 1987
onwards.
Beginning with
OBRA
'86, states
were
given the option, and then mandated
regardless of family structure. OBRA '86 gave
flrst
to, increase the
income thresholds for Medicaid eligibility,
states the option
women up to 100% of the poverty threshold, beginning in April, 1987. OBRA '87
increased that optional level to 185% of poverty. Under the Medicare Catastrophic Care Act states were
of covering pregnant
mandated
to
cover pregnant
required to cover
women up
women up to 133%
to
75%
Then under OBRA '89, they were
1990. Using data from Government Accounting
of poverty by July
of poverty by April,
1.
Office (1991), Fischer (1992), and Intergovernmental Health Policy Project (various years),
we modelled
women
was covered
as being covered if their
income was below the average percentage of poverty
during the year.
44
that
k
dS6
5 7
MIT LIBRARIES
3
^DflD
D0a4bS^3
T
Date Due
^t.
1
WW
MAY
Lib-26-67