Has Public Health Insurance for Older Has Public Health Insurance for Older Children Reduced Disparities in Access Children Reduced Disparities in Access

to Care and Health Outcomes?to Care and Health Outcomes?

Janet Currie,Janet Currie, Sandra Decker, and Sandra Decker, and Wanchuan LinWanchuan Lin

AcademyHealthAcademyHealthWashington D.C.Washington D.C.

June 10, 2008June 10, 2008

DisclaimerDisclaimer

The findings and conclusions in this presentation areThe findings and conclusions in this presentation are

those of the authors and do not necessarily represent those of the authors and do not necessarily represent

the views of the National Center for Health Statistics the views of the National Center for Health Statistics

or the Centers for Disease Control and Prevention.or the Centers for Disease Control and Prevention.

BackgroundBackground

• • Children of lower income parents are in worse health Children of lower income parents are in worse health compared to other children. compared to other children.

•• This fact is even more true for older children compared to This fact is even more true for older children compared to younger (Case, Lubotsky, and Paxson (CLP), younger (Case, Lubotsky, and Paxson (CLP), American American Economic ReviewEconomic Review, 2002) , 2002)

•• Since nearly one in five U.S. children live in poor families, Since nearly one in five U.S. children live in poor families, it is important to understand how low income results in it is important to understand how low income results in poor health and why this relationship is even stronger for poor health and why this relationship is even stronger for older children than younger.older children than younger.

Question #1: Question #1:

Has the relationship between Has the relationship between income and child health changed income and child health changed over time? Has it changed over time? Has it changed differently for children of different differently for children of different ages?ages?

DataData

National Health Interview Survey (NHIS), 1986-National Health Interview Survey (NHIS), 1986-20052005

After dropping children missing information on key variables such After dropping children missing information on key variables such as race or health status and children whose age and recorded birth as race or health status and children whose age and recorded birth month and year don’t match, month and year don’t match, our final sample size consists of our final sample size consists of 475,517 children under age 18475,517 children under age 18 from NHIS 1986-2005. from NHIS 1986-2005.

We use NCHS income imputations for 1990-2005, and perform our We use NCHS income imputations for 1990-2005, and perform our own income imputations for 1986-1989 using a method similar to own income imputations for 1986-1989 using a method similar to the method used by NCHS for 1990-1996.the method used by NCHS for 1990-1996.

Following CLP, we assign precise incomes to income brackets (by Following CLP, we assign precise incomes to income brackets (by education of reference person or spouse) using the March Current education of reference person or spouse) using the March Current Population Survey (CPS) 1986-2005.Population Survey (CPS) 1986-2005.

Ages 0-3 Ages 4-8 Ages 9-12 Ages 13-17

Log Family Income ($1986) -0.136 -0.152 -0.194 -0.220

[0.011]*** [0.011]*** [0.010]*** [0.011]***

* 1991-1995 Time Period -0.002 -0.011 0.005 0.013

[0.013] [0.013] [0.014] [0.015]

* 1996-2000 Time Period -0.011 -0.025 0.006 0.041

[0.014] [0.013] [0.013] [0.014]***

* 2000-2005 Time Period 0.015 0.018 0.034 0.051

[0.013] [0.013] [0.013]*** [0.013]***

Observations 101,386 136,677 109,033 128,420

The Effect of Income on Child Health

The table reports coefficients and standard errors (in brackets) from ordered probit models where the dependent variable is child self-reported health. Although not reported, controls include year, state and age effects, as well as (log of) family size, whether mother and father present, race (white, black), mother's age, father's age, mother's education (less than 12 or 12 years), father's education (less than 12 or 12 years), and whether the mother or father is unemployed. * significant at 10%, ** significant at 5%, ***significant at 1%.

Ordered Probit (1=Excellent, 2=Very Good, 3=Good, 4=Fair, 5=Poor)

Ages 0-3 Ages 4-8 Ages 9-12 Ages 13-17

Log Family Income ($1986) -0.046 -0.052 -0.066 -0.076

[0.004]*** [0.004]*** [0.004]*** [0.004]***

* 1991-1995 Time Period -0.004 -0.005 0.001 0.005

[0.005] [0.005] [0.005] [0.006]

* 1996-2000 Time Period -0.007 -0.011 0.002 0.014

[0.006] [0.006]** [0.006] [0.006]**

* 2000-2005 Time Period 0.005 0.005 0.011 0.016

[0.006] [0.005] [0.006]* [0.006]***

Observations 101,386 136,677 109,033 128,420

The Effect of Income on Child Health

The table reports coefficients and standard errors (in brackets) from linear probability models estimating the probability that a child is in less than excellent health. Although not reported,

controls include year, state and age effects as well as (log of) family size, whether mother and father present, race (white, black), mother's age, father's age, mother's education (less than 12 or 12 years), father's education (less than 12 or 12 years), and whether the mother or father is unemployed. * significant at 10%, ** significant at 5%, ***significant at 1%.

Linear Probability (Dependent Variable: Less Than Excellent Health)

Question # 2:Question # 2:

What explains the decline in the What explains the decline in the relationship between income relationship between income and health for older children?and health for older children?

Medicaid/SCHIP EligibilityMedicaid/SCHIP Eligibility

1.1. Historically, a child’s eligibility for Medicaid was tied to the receipt Historically, a child’s eligibility for Medicaid was tied to the receipt of cash welfare payments under the Aid for Families with of cash welfare payments under the Aid for Families with Dependent Children (AFDC) program. Dependent Children (AFDC) program.

2.2. Beginning in 1984, Congress expanded Medicaid coverage to Beginning in 1984, Congress expanded Medicaid coverage to pregnant women, infants and younger children not on welfare. By pregnant women, infants and younger children not on welfare. By April 1, 1990, states were required to offer coverage to all children April 1, 1990, states were required to offer coverage to all children up to age six in families with income up to 133% of the Federal up to age six in families with income up to 133% of the Federal poverty level. poverty level.

3.3. OBRA 1990 required states to increase eligibility of older children OBRA 1990 required states to increase eligibility of older children under the poverty level by one more year of age per year for under the poverty level by one more year of age per year for children born after September 1983. By 2001, all poor children children born after September 1983. By 2001, all poor children were required to be made eligible. were required to be made eligible.

4.4. SCHIP initiated in 1996 provided an additional source of public SCHIP initiated in 1996 provided an additional source of public health insurance for low income children.health insurance for low income children.

Empirical SpecificationEmpirical Specification

….”other controls” – X – include age dummies, (log of) family size, whether mother and father present, race (white, black), mother's age, father's age, mother's education (less than 12 or 12 years), father's education (less than 12 or 12 years), and whether the mother or father is unemployed.

(2) y = b0 + b 1PUBINS + b 2INC + b 3INC*AGE9 to 17 +

b4 INC*T_1996-2005 + b 5 AGE9 to 17*T_1996-2005 +

b 6INC*AGE9 to 17*T_1996 -2005+ b 7 X + b8 STATE + b9 YEAR +

b10 STATE*AGE 9 to 17 + e,

where PUBINS indicates that the child is eligible for public health

insurance, AGE9 to 17 indicates that they are aged 9 to 17 and

T_1996-2005 indicates that it is the second half of our time period.

Endogeneity of PUBINSEndogeneity of PUBINS

Persons who are eligible for public health Persons who are eligible for public health insurance may be different from others in insurance may be different from others in unobserved waysunobserved ways

For example, a sick child may cause lower For example, a sick child may cause lower parental income, leading to a spurious parental income, leading to a spurious correlation between public health insurance correlation between public health insurance eligibility and poor health.eligibility and poor health.

We therefore use a two-staged instrumental variables We therefore use a two-staged instrumental variables approach to estimate the effect of Medicaid/SCHIP approach to estimate the effect of Medicaid/SCHIP eligibility (PUBLINS) on child healtheligibility (PUBLINS) on child health

Simulated Medicaid/SCHIP EligibilitySimulated Medicaid/SCHIP Eligibility We generate Medicaid/SCHIP simulated eligibility by We generate Medicaid/SCHIP simulated eligibility by

sampling 500 children by single year of age and calendar year sampling 500 children by single year of age and calendar year from the CPS, and then calculating the fraction of this fixed from the CPS, and then calculating the fraction of this fixed group of children who would be eligible for Medicaid/SCHIP group of children who would be eligible for Medicaid/SCHIP in each state and year. in each state and year.

Instrument for individual eligibility using the simulated Instrument for individual eligibility using the simulated eligibilityeligibility..

GoalGoal: Abstract from characteristics of the child and or : Abstract from characteristics of the child and or family that may be correlated with both eligibility and the family that may be correlated with both eligibility and the dependent variables to achieve identification using only dependent variables to achieve identification using only legislative variation in Medicaid/SCHIP policy.legislative variation in Medicaid/SCHIP policy.

0

50

100

150

200

250

Me

dic

aid

/SC

HIP

Elig

ibili

ty C

uto

ffa

s a

Fra

ctio

n o

f th

e F

ed

era

l Po

vert

y L

eve

l

19

86

19

88

19

90

19

92

19

94

19

96

19

98

20

00

20

02

20

04

year

Ages 0-3 Ages 4-8

Ages 9-12 Ages 13~17

TEXAS: Medicaid/SCHIP Eligibility by Child Age Group

0

50

100

150

200

250

Me

dic

aid

/SC

HIP

Elig

ibili

ty C

uto

ffa

s a

Fra

ctio

n o

f th

e F

ed

era

l Po

vert

y L

eve

l

19

86

19

88

19

90

19

92

19

94

19

96

19

98

20

00

20

02

20

04

year

Ages 0-3 Ages 4-8

Ages 9-12 Ages 13~17

CALIFORNIA: Medicaid/SCHIP Eligibility by Child Age Group

Dependent Variable:

OLS OLS TSLS

Medicaid/SCHIP Eligible 0.024 -0.005

[0.004]*** [0.020]

Log Family Income ($1986) -0.049 -0.041 -0.05

[0.002]*** [0.003]*** [0.006]***

Log Family Income ($1986) * Ages 9-17 -0.022 -0.023 -0.022

[0.003]*** [0.003]*** [0.003]***

Log Family Income ($1986) * 1996-2005 0.001 0.001 0.001

[0.003] [0.003] [0.003]

Ages 9-17 * 1996-2005 -0.055 -0.071 -0.052

[0.040] [0.040]* [0.042]

Log Family Income ($1986) * Ages 9-17 * 1996-2005 0.009 0.011 0.009

[0.004]** [0.004]*** [0.004]**

First Stage F-statistic 4,125

P-value for first stage F-statistic 0.000

The Effect of Medicaid/SCHIP Eligibility on Child Health

The table reports coefficients and standard errors (in brackets) from linear probability models estimated using ordinary least squares (OLS) and two staged least squares (TSLS). Other than those indicated in the table, control variables include age effects, year effects, state effects, state effects interacted with age group, and (log of) family size, whether mother and father present, race (white, black), mother's age, father's age, mother's education (less than 12 or 12 years), father's education (less than 12 or 12 years), and whether the mother or father is unemployed. The sample size is 475,516 for less than excellent health and 378,233 for no doctor visit in the past year. * significant at 10%, ** significant at 5%, ***significant at 1%.

Less Than Excellent Health

Linear Probability Models

17

Question #3:Question #3:

Does Medicaid/SCHIP eligibility in Does Medicaid/SCHIP eligibility in early childhood have positive early childhood have positive future effects on child health?future effects on child health?

Lagged Medicaid/SCHIP EligbilityLagged Medicaid/SCHIP Eligbility

Since health is a stock, it is affected by past Since health is a stock, it is affected by past investments as well as current ones.investments as well as current ones.

It may not be very surprising that It may not be very surprising that contemporaneous Medicaid/SCHIP eligibility contemporaneous Medicaid/SCHIP eligibility has little effect on overall health status.has little effect on overall health status.

Therefore, we investigate whether Therefore, we investigate whether Medicaid/SCHIP eligibility in early childhood Medicaid/SCHIP eligibility in early childhood puts children on a better health trajectory, puts children on a better health trajectory, resulting in better health at older ages. resulting in better health at older ages.

Empirical SpecificationEmpirical Specification…reduced form models……reduced form models…

….”other controls” – X – include age dummies, (log of) family size, whether mother and father present, race (white, black), mother's age, father's age, mother's education (less than 12 or 12 years), father's education (less than 12 or 12 years), and whether the mother or father is unemployed.

Because we are interested in the effect of lagged health insurance eligibility

on the health status of older children, we estimate these models using only the

children aged 9 -17. These models are of the form:

(3) y = b0 + b1PUBINS + b2INC + b3INC*T_1996-2000 + b4X + b5STATE +

b6YEAR + e,

where PUBINS is now a measure of the fraction eligible for public health

insurance when the child was age 0, age 1, age 2, etc. Eligibility at each age is

included in a separate regression since there is a good deal of multicollinearity

between eligibility at various ages.

Reduced Form Models

Controlling for Simulated Eligibility at: Age 0 Age 1 Age 2 Age 3 Age 4 Age 5 Age 6 Age 7 Age 8

Dependent Variable: Less Than Excellent Health

Lagged Simulated Eligible -0.021 -0.037 -0.038 -0.049 -0.041 -0.034 -0.026 -0.016 -0.015[0.024] [0.024] [0.022]* [0.024]** [0.023]* [0.022] [0.020] [0.020] [0.020]

Log Family Income ($1986) -0.069 -0.069 -0.069 -0.069 -0.069 -0.069 -0.069 -0.069 -0.069[0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]***

Log Family Income ($1986) * 1996-2005 0.009 0.009 0.009 0.009 0.009 0.009 0.01 0.01 0.01[0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]***

The Effect of Medicaid/SCHIP Eligibility and Lagged Medicaid/SCHIP Eligibility on Health of Children Aged 9-17

The table reports coefficients and standard errors (in brackets) from linear probability models. Other than those indicated in the table, control variables include state, year and age effects, and (log of) family size, whether mother and father present, race (white, black), mother's age, father's age, mother's education (lesss than 12 or 12 years), father's education (less than 12 or 12 years), and whether the mother or father is unemployed. The sample size is 237,453. * significant at 10%, ** significant at 5%, ***significant at 1%.

Linear Probability Models

Reduced Form Models

Controlling for Simulated Eligibility at: Age 0 Age 1 Age 2 Age 3 Age 4 Age 5 Age 6 Age 7 Age 8

Dependent Variable: No Doctor Visit in the Past Year

Lagged Simulated Eligible -0.073 -0.09 -0.09 -0.066 -0.065 -0.054 -0.05 -0.042 -0.053[0.020]*** [0.023]*** [0.022]*** [0.022]*** [0.022]*** [0.022]** [0.020]** [0.020]** [0.019]***

Log Family Income ($1986) -0.032 -0.032 -0.032 -0.033 -0.033 -0.033 -0.033 -0.033 -0.033[0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]*** [0.003]***

Log Family Income ($1986) * 1996-2005 0.007 0.007 0.007 0.007 0.007 0.007 0.007 0.007 0.007[0.003]** [0.003]** [0.003]** [0.003]** [0.003]** [0.003]** [0.003]** [0.003]** [0.003]**

The Effect of Medicaid/SCHIP Eligibility and Lagged Medicaid/SCHIP Eligibility on Access to Health Care of Children Aged 9-17

The table reports coefficients and standard errors (in brackets) from linear probability models. Other than those indicated in the table, control variables include state, year and age effects, and (log of) family size, whether mother and father present, race (white, black), mother's age, father's age, mother's education (lesss than 12 or 12 years), father's education (less than 12 or 12 years), and whether the mother or father is unemployed. The sample size is 188,299. * significant at 10%, ** significant at 5%, ***significant at 1%.

Linear Probability Models

ConclusionsConclusions In 1996-2005, the relationship between income and In 1996-2005, the relationship between income and

health continues to be stronger for older children health continues to be stronger for older children compared to younger. compared to younger.

However, the gradient increase with age is less in 1996-However, the gradient increase with age is less in 1996-2005 than it was in 1986-1995. 2005 than it was in 1986-1995.

Increases in Medicaid/SCHIP eligibility increase the Increases in Medicaid/SCHIP eligibility increase the probability that a child has seen a doctor in the past probability that a child has seen a doctor in the past year.year.

We see no evidence that this improvement in the use of We see no evidence that this improvement in the use of health services can explain the decline in the strength of health services can explain the decline in the strength of the relationship between income and health for older the relationship between income and health for older children. children.

We find some evidence that Medicaid/SCHIP eligibility We find some evidence that Medicaid/SCHIP eligibility in early childhood puts children on a better health in early childhood puts children on a better health trajectory, resulting in better health at older ages.trajectory, resulting in better health at older ages.