Expanding Public Health Insurance to Parents: Effects on Children's Coverage under Medicaid

Abstract

Objective

To assess whether expanding public health insurance coverage to parents leads to increases in Medicaid participation among children.

Data Sources/Study Setting

Study uses data from the 1997 and 1999 National Survey of America's Families. Insurance coverage of children eligible for Medicaid under the poverty-related expansions is analyzed.

Study Design

We conduct two analyses. In the first, we examine the cross-sectional difference regarding whether Medicaid participation is higher for children eligible for Medicaid under the poverty-related expansions when states expand public health insurance programs to cover their parents. In the second, we use a difference in-difference approach to assess whether the expansion of the Medicaid program to cover parents in Massachusetts led to an increase in Medicaid coverage among children between 1997 and 1999 relative to changes that occurred in the rest of the nation.

Data Collection/Extraction Methods

The analysis relies on a detailed Medicaid and SCHIP eligibility simulation model that identifies children surveyed on the NSAF who are eligible for Medicaid under the poverty-related expansions.

Principal Findings

Children who reside in states that expanded public health insurance programs to parents participate in Medicaid at a rate that is 20 percentage points higher than of those who live in states with no expansions. The Massachusetts expansion in coverage to parents led to a 14 percentage point increase in Medicaid coverage among children due principally to reductions in uninsurance among already eligible children.

Conclusions

Expanding public health insurance coverage to parents has benefits to children in the form of increased participation in Medicaid.

Over the past several years there as been increasing policy interest in expanding eligibility for public health insurance coverage to parents. Recent changes in federal policy have provided greater access to federal matching dollars to states that want to expand coverage to parents under Medicaid and the State Children's Health Insurance Program (SCHIP). These changes include the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA), which delinked eligibility for Medicaid coverage from eligibility for cash assistance, SCHIP, and Health Insurance Flexibility and Accountability (HIFA) waivers (Dubay, Kenney, and Zuckerman 2000; Dubay and Kenney 2001a; Howell et al. 2002). A number of states have taken advantage of these new options under Medicaid and SCHIP and others have developed separate programs using state funds to cover parents (Krebs-Carter and Holahan 2000; Howell et al. 2002).

While such coverage expansions have clear potential benefits to parents (Dubay and Kenney 2001a; Kronick and Gilmer 2002), there is emerging evidence that extending coverage to parents may also benefit children. Ku and Broaddus (2000) find that between 1990 and 1998 Medicaid coverage of children under age 6 with incomes below 133 percent of the federal poverty level (FPL) increased more in states with broad expansions in coverage to parents relative to states that had no parent expansions. Lambrew (2001) notes that the uninsurance rate for children is lower in states that have implemented expansions in coverage to parents. Finally, Davidoff et al. (forthcoming) and Gifford et al. (2001) show that insured children with uninsured parents are less likely to receive well-child visits than those with an insured parent.

In this paper we assess the extent to which covering parents increases participation of children eligible for Medicaid. We extend the previous literature by using a detailed Medicaid eligibility simulation model that allows us to identify children eligible for Medicaid under the poverty-related expansions. We focus on these children because their parents have not historically been eligible for public coverage unless the state had a waiver or a state-funded program.1 In addition, parents of these children have been the target of recent expansions in the Medicaid program. We compare the experience of children eligible for Medicaid under the poverty-related expansions in states that have expanded health insurance coverage to their parents to that of similarly situated children in states with no parental coverage expansions to assess whether extending coverage to parents has the spillover effect of increasing participation of children in Medicaid. Results from two separate analyses indicate that extending eligibility for insurance coverage to parents increases participation in Medicaid among children and leads to lower rates of uninsurance. By extending eligibility to more parents, states may increase coverage for both parents and children. However, we do find that some of the increased participation may be due to families substituting public coverage for the private coverage they used to have, but we find that the magnitude of this apparent substitution is low.

BACKGROUND ON MEDICAID AND SCHIP COVERAGE POLICIES

Until the mid-1980s, Medicaid coverage for children was limited primarily to children living in families that qualified for Aid to Families with Dependent Children (AFDC). Beginning with the Medicare Catastrophic Coverage Act (MCCA) of 1988, a series of eligibility expansions were mandated for children, called poverty-related expansions. Ultimately, states were required to cover children under age 6 in families with incomes up to 133 percent of poverty, and children born after September 30, 1983, in families with incomes up to 100 percent of poverty. In addition to these mandates, a number of states took advantage of options through Section 1902(r)(2) provisions and Section 1115 waivers in the early to middle 1990s to further expand coverage to children. Following the creation of SCHIP in 1997, additional eligibility expansions occurred for children in all states. As a consequence, all but 16 percent of low-income (defined as below 200 percent FPL) uninsured children are now eligible for coverage under either Title XIX or Title XXI (Dubay, Haley, and Kenney 2002).

Coverage expansions for parents have not kept pace with the expansions for children. Coverage of nonelderly adults under Medicaid has historically been limited to parents receiving cash assistance under Aid to Families with Dependent Children (AFDC), disabled adults receiving Supplemental Security Income (SSI), and, since the mid-1980s, pregnant women. Many poor and near-poor parents were ineligible for Medicaid because AFDC eligibility was restricted to very low income single-parent and two-parent families where either one parent was incapacitated or the principal wage earner was unemployed.2

Four federal changes dramatically expanded the options available to states for covering low-income parents under Medicaid or SCHIP.3 First, PRWORA in 1996 created a new category of Medicaid eligibility in Section 1931 of the Social Security Act. It requires states to grant such eligibility to those adults and children who would have been entitled to AFDC under the income and resource standards in effect on July 16, 1996.4 It also gives states the option to use less restrictive methodologies for counting income and resources when determining eligibility—thus allowing states to make higher income families that meet the categorical requirements under the old AFDC program eligible for Medicaid.5 Second, the Department of Health and Human Services issued a regulation in August 1998 that permits states to use less-restrictive rules in defining unemployment for two-parent families, essentially allowing states to cover all two-parent families that meet the Section 1931 income and resource requirements.6 Importantly, Section 1931 eligibility provisions apply to families, making it impossible for parents to be made eligible without their children being eligible.

The third major federal change came in July 2000, when HCFA issued guidance to states regarding the use of SCHIP waivers to cover low-income parents under SCHIP. To obtain a waiver to cover parents under SCHIP, states must cover children up to 200 percent of poverty, enroll children statewide without any waiting list, demonstrate that the application and redetermination processes for Medicaid and SCHIP promote enrollment and retention of children in the programs, and make lower income parents eligible for coverage prior to making higher income parents eligible. Under these waivers, states that implemented expansions for parents prior to March 31, 2000, will continue to receive the Medicaid matching rate for parents with incomes below 100 percent of the FPL and will get the higher SCHIP match for parents with higher incomes. States that implement expansions to parents after March 31, 2000, will receive the enhanced SCHIP match for all parents with incomes above the threshold they had in place prior to the expansion.7 Under both of these circumstances, the enhanced SCHIP match will only be available if unspent SCHIP funds exist for a given state.

Finally, in August of 2001, the Department of Health and Human Services issued guidelines for waivers under the Health Insurance Flexibility and Accountability (HIFA) initiative. Under these waivers states can vary benefit packages to groups that they are not mandated to cover and use their SCHIP allocations to cover other populations. To date, six states, Arizona, California, Illinois, Maine, New Mexico, and Oregon have the authority to use this mechanism to expand coverage to parents.8

The following summarizes the country's progress in state initiatives to expand coverage of low-income parents beyond the welfare thresholds as of 1997 and 1999, the two years examined in our study. In 1997, five states (Delaware, Hawaii, Massachusetts, Tennessee, and Vermont) had expanded coverage to parents solely through their Medicaid program; two (Minnesota and Washington) had expanded coverage to low-income parents solely through a state-funded program; and one (Oregon) had expanded coverage to low-income parents though both routes. By 1999, four additional states (District of Columbia, Missouri, Rhode Island, and Wisconsin) had expanded coverage to low-income parents through their Medicaid program.

DATA

This analysis draws on the 1997 and 1999 waves of the National Survey of America's Families (NSAF). The NSAF is a household survey that provides information on more than 100,000 children and adults representing the noninstitutionalized civilian population under age 65. The NSAF oversamples the low-income population (defined as having incomes below 200 percent of the FPL, $33,400 for a family of four in 1999) and provides nationally representative estimates as well as state-representative estimates for 13 states.9 These 13 states were selected for intensive study because they represented a mixture of approaches to health and social policy and because they were diverse geographically and economically.

Four of the 13 NSAF focal states (Massachusetts, Minnesota, Washington, and Wisconsin) have expanded coverage to parents beyond welfare thresholds by the end of 1999. Since Massachusetts did not introduce its 1997 initiative until July, which was very late in the first NSAF survey period, we were able to do a special analysis comparing Massachusetts preexpansion in 1997 with Massachusetts postexpansion in 1999. (We cannot do the same with Wisconsin, because its introduction of parent coverage was similarly late in the second NSAF survey period.)

NSAF interviews were conducted in households with and without telephones using computer-assisted telephone interviewing technology.10 The data are weighted to provide reliable national and state estimates. The standard errors are based on the balanced repeated replication method to account for the complex nature of the sample design (Brick et al. 1999). Detailed information was collected from the adult (called the most knowledgeable adult, or MKA) who knew most about the education and health care of up to two children (one age 5 and under and one from ages 6 to 17) in each selected household. For this analysis, parents are defined to include biological, adoptive, or stepparents, as long as they are living in the household of the child.11

Current insurance coverage is measured through a series of questions on coverage at the time of the survey.12 Coverage includes private employer-sponsored and nongroup plans, as well as Medicaid, SCHIP, other state programs, Medicare, and other public programs such as CHAMPUS. State-specific program names were inserted in these questions to enhance respondents' recognition of programs, and we added a new question to the 1999 instrument asking about separate SCHIP programs. Because more than one type of coverage was reported for a small number of children, a hierarchy was developed to classify people into mutually exclusive groups. Coverage through Medicaid, SCHIP, or another state program took precedence, followed by employer-sponsored and nongroup plans, and then any other insurance coverage.13

The analyses presented in this paper rely on a detailed Medicaid and SCHIP eligibility simulation model designed to mimic the eligibility determination process faced by families applying for Medicaid or SCHIP.14 First, eligibility units are created from the household survey data. Individuals included in the unit are only those who would be considered in the eligibility determination process.15 Second, Medicaid and SCHIP eligibility rules in place in July of each year are applied to each unit. Relevant rules include those regarding: eligibility thresholds, which vary by age of the child, family composition, and work status of the parents; how income is counted, including whose income is counted and what types of unearned income are counted; work, earned income, child care, and child support disregards; asset limits and disregards; and deeming of stepparent and grandparent income. Third, children are categorized into three eligibility groups hierarchically: (a) those who are eligible for Medicaid but would also have been eligible for TANF (TANF-related);16, 17 (b) those who are eligible for Medicaid based on the poverty-related expansions, both those federally mandated and those allowed under Section 1902(r)(2) provisions, and Section 1115 waiver authority (poverty-related)18; and (c) those who are eligible for SCHIP whether through expanded Medicaid or through separate programs created under SCHIP.

METHODS

We use two different approaches to address whether extending public health insurance coverage to parents leads to greater participation in Medicaid among children who are already eligible. In the first, we contrast Medicaid participation rates in 1999 for children eligible for Medicaid under the poverty-related expansions in states that have expanded coverage to parents to participation rates for children in states that do not cover parents. In the second, we take advantage of the fact that Massachusetts, one of thirteen states oversampled in the NSAF, implemented its family coverage expansion after the first round and before the second round of the NSAF.19

Comparison of State Groups

In this analysis we focus on children eligible under the poverty-related expansions prior to welfare reform, for two reasons. First, the parents of these children were not eligible for public coverage unless the state had a Section 1115 waiver under Medicaid or had a state-funded program.20 Second, these are the parents targeted by the recent policies to expand family coverage. Parents of children eligible under the TANF-related rules, in contrast, have always been eligible for Medicaid. We make separate 1999 estimates for states that had expanded Medicaid to parents under either Section 1115 waivers or through Section 1931 provisions (Delaware, District of Columbia, Hawaii, Massachusetts, Missouri, Oregon, Rhode Island, Tennessee, and Vermont) and for states that created state-funded, non-Medicaid, programs to cover parents (Minnesota and Washington).

We exclude from the participation rate calculation children with private insurance coverage because we want to measure the extent to which the Medicaid program is reaching the eligible but uninsured population.21 This approach allows us to account for the variation in private insurance coverage across states by examining the extent to which public programs close the gap in coverage left by private insurance (Spillman 2000). We present both descriptive and multivariate results. In the multivariate models, we estimate the following equation for children eligible for Medicaid under the poverty-related expansions:

$$Medicaid\hspace{0.17em}=\hspace{0.17em}{\beta }_0\hspace{0.17em}+\hspace{0.17em}{\beta }_{1}fcsp\hspace{0.17em}+\hspace{0.17em}{\beta }_{2}fcmp\hspace{0.17em}+\hspace{0.17em}{\beta }_{3}child\hspace{0.17em}+\hspace{0.17em}{\beta }_{4}parent\hspace{0.17em}+\hspace{0.17em}{\beta }_{5}family$$ (1)

Where:

Medicaid=1 if the child eligible under the poverty-related expansions participates in Medicaid and 0 if child is uninsured

fcsp=1 if the state covers parents under a separate program

fcmp=1 if the state covers parents under the Medicaid program

child=a vector of child characteristics

parent=a vector of parent characteristics

family=a vector of family characteristics

In the context of this regression, the coefficients on fcsp and fcmp represent the difference in participation rates between children in states with no family coverage and with the two types of family coverage being analyzed. Child characteristics include: age, race, and health status of the child; parent characteristics include: their work status, education, and nativity; family characteristics include: income, welfare history, and number of children.22

It is important to note that differences in Medicaid participation that are observed between states with and without family coverage in this type of cross-sectional analysis may be due to unmeasured differences between the two groups of states in factors such as program quality, awareness of the program, or ease of enrollment. We conduct a number of additional analyses to assess the extent to which these results appear to be attributable to unmeasured differences rather than to differences in family coverage policies. Specifically, we examine whether similar differences exist across these states in Medicaid participation among children eligible for Medicaid under the TANF-related rules, by reestimating equation 1 just for children eligible under the TANF-related rules. If we find that participation of children eligible under the TANF-related rules does not vary between the states with and without family coverage to the same extent as for children under the poverty-related expansions, we would feel more confident in attributing the participation differences to family coverage policies.

We also examine whether the difference in participation rates between children eligible under the poverty-related expansions and under TANF-related rules vary by whether the state has expanded family coverage. To the extent that we find similar differentials between these two rates in the different groups of states, we will conclude that unmeasured differences across state groups are influencing the participation rates, not the family coverage policies. We estimate equation 2 separately for each of the three groups of states:

$$\begin{array}{c}\mathrm{Medicaid}=\hspace{0.17em}1\text{ if the child in state group }i\text{ participates in Medicaid and }0\\ \text{ if child is uninsured}\\ \mathrm{Pov}\hspace{0.17em}=\hspace{0.17em}1\text{ if child is eligible for Medicaid under the poverty-related expansions and }\\ 0\text{ if child is eligible for Medicaid under TANF-related rules}.\end{array}$$ 2

Using this specification, the coefficient on Pov represents the difference in participation rates between children eligible under TANF- and poverty-related rules in each of the groups of states, and child, parent, and family are defined as before.

Finally, we net out underlying differences in participation across the different types of states and then assess whether children eligible under the poverty-related expansions participate in Medicaid at higher rates when the public program offers coverage to their parents. The effects of family coverage on Medicaid participation rates net of state-specific participation effects are estimated using equation 3

$$\mathrm{Medicaid}\hspace{0.17em}=\hspace{0.17em}{\beta }_0\hspace{0.17em}+\hspace{0.17em}{\beta }_{1}pov\hspace{0.17em}+\hspace{0.17em}{\beta }_{2}fcsp\hspace{0.17em}+\hspace{0.17em}{\beta }_{3}fcmp\hspace{0.17em}+\hspace{0.17em}{\beta }_{4}fcs{p}^{*}pov\hspace{0.17em}+\hspace{0.17em}{\beta }_{5}fcs{p}^{*}pov\hspace{0.17em}+\hspace{0.17em}{\beta }_{6}child\hspace{0.17em}+{\beta }_{7}Parent\hspace{0.17em}+{\beta }_{8}Family$$ (3)

Using this specification, the coefficients on fcsp^*pov and fcmp^*pov represent the increase in Medicaid participation that occurs in states with family coverage net of underlying differences in participation across the states. All the other variables are defined as before.

The Massachusetts Before and After Comparison

In this analysis we contrast changes in Medicaid coverage for children in Massachusetts before and after implementation of family coverage with changes in Medicaid coverage for children over this period in the rest of the nation. We also contrast changes in private coverage and in the uninsurance rate in Massachusetts with changes in the rest of the nation in order to assess whether the observed increases in coverage are due to reductions in the uninsurance rate or the substitution of public for private coverage. This difference-in-difference approach explicitly uses trends in insurance coverage for the rest of the nation as a control for what would have happened in Massachusetts in the absence of the family coverage expansions.23 This methodology has been used extensively to examine the impact of previous Medicaid expansions for children on insurance coverage.24

We focus on children who were already eligible for Medicaid under the poverty-related expansions but whose parents were made newly eligible. We conduct both descriptive and multivariate analyses.25,26 We estimate the effects of family coverage in Massachusetts on insurance coverage of children using equation 4–equation 6

$$\text{Medicaid}\hspace{0.17em}=\hspace{0.17em}{\beta }_0\hspace{0.17em}+\hspace{0.17em}{\beta }_{1}Mass\hspace{0.17em}+\hspace{0.17em}{\beta }_{2}Year\hspace{0.17em}+\hspace{0.17em}{\beta }_{3}Mas{s}^{*}Year\hspace{0.17em}+\hspace{0.17em}{\beta }_4\hspace{0.17em}Child\hspace{0.17em}+\hspace{0.17em}{\beta }_{5}Parent\hspace{0.17em}+{\beta }_{6}Family$$ (4)

$$Private\hspace{0.17em}=\hspace{0.17em}{\delta }_0\hspace{0.17em}+\hspace{0.17em}{\delta }_{1}Mass\hspace{0.17em}+\hspace{0.17em}{\delta }_{2}Year\hspace{0.17em}+\hspace{0.17em}{\delta }_{3}Mas{s}^{*}Year\hspace{0.17em}+\hspace{0.17em}{\delta }_{4}Child\hspace{0.17em}+\hspace{0.17em}{\delta }_{5}Parent\hspace{0.17em}+\hspace{0.17em}{\delta }_{6}Family$$ (5)

$$\text{Unisured}\hspace{0.17em}=\hspace{0.17em}{\lambda }_0\hspace{0.17em}+\hspace{0.17em}{\lambda }_{1}Mass\hspace{0.17em}+\hspace{0.17em}{\lambda }_{2}Year\hspace{0.17em}+\hspace{0.17em}{\lambda }_{3}Mas{s}^{*}Year\hspace{0.17em}+\hspace{0.17em}{\lambda }_{4}Child\hspace{0.17em}+\hspace{0.17em}{\lambda }_{5}Parent\hspace{0.17em}+\hspace{0.17em}{\lambda }_{6}Family$$ (6)

Where:

Medicaid=1 if child eligible under poverty-related expansions participates in Medicaid

Private=1 if child eligible under the poverty-related expansions has private coverage

Uninsured=1 if child eligible under the poverty-related expansions is uninsured

Mass=1 if child resides in Massachusetts

Year=1 if year is 1999

Using this specification, the coefficients on Mass^*year in each equation represent the difference-in-difference estimate of the effect of family coverage on insurance coverage. The share of the increase in Medicaid participation attributable to reductions in uninsurance is the ratio of λ₃ from the model predicting uninsurance to β₃ from the model predicting Medicaid participation.

The multivariate analysis includes the same child, parent, and family control variables used in the comparison of state groups model with one exception. We conduct this analysis with and without controls for offers of employer-sponsored insurance (ESI) coverage. We exclude offers of coverage because, in theory, the offer rate could be affected by the broad-based expansion of coverage in Massachusetts through the MassHealth Program, making inclusion as a control variable endogenous. At the same time, family premiums per enrolled employee increased by 13.6 percent nationally over this period (Agency for Health Care Research and Quality 2000, 2001). This trend in premiums may have affected employers' willingness to offer coverage and individuals' willingness to take up ESI even absent the coverage expansion. Importantly, analyses that have examined the issue of whether employers responded to the Medicaid expansions for children and pregnant women have found no evidence of an employer response (Cutler and Gruber 1996; Shore-Sheppard, Buchmueller, and Jensen 2000). Estimating separate models that both exclude and include a variable measuring offers of ESI coverage yields upper- and lower-bound estimates of the extent to which any increase in Medicaid coverage was due to the substitution of private coverage. Given the lack of evidence of an employer response under previous expansions, we give greater emphasis to the results that control for offers of ESI.

RESULTS

Comparison of State Groups

Table 1 presents Medicaid participation rates for children eligible for Medicaid under the poverty-related expansions for states with different family coverage policies.27 These results suggest that expanding coverage to parents leads to greater Medicaid participation among children who are already eligible. In states without family coverage expansions, 57.1 percent of poverty-related children participate in the Medicaid program, compared to 78.5 percent in states that have state-funded family coverage expansions and 80.8 percent in states that have Medicaid family coverage expansions. When we control for differences across states in the characteristics of the eligible population, we still find that Medicaid participation rates among poverty-related children in states with family coverage expansions are about 24 percentage points higher than in states with no family coverage.

Table 1.

Medicaid Participation Rates for Children Eligible under the Poverty-Related Expansions by Family Coverage Status, 1999¹

		Difference2	Regression-Adjusted Difference2,3
No family coverage	57.1
Family coverage under state-funded program4	78.5*	21.4*	21.9*
Family coverage under Medicaid program5	80.8*	23.6*	24.0*

¹Participation rates exclude children with private insurance coverage and Medicare-only coverage.

²Differences are measured between states with no family coverage and states with family coverage.

³Regression models control for age, race, and health status of child; education, work status, and nativity of parents; and income, number of children, and welfare history of the family.

⁴Includes Minnesota and Washington.

⁵Includes Oregon, Vermont, Delaware, Hawaii, Tennessee, Missouri, Massachusetts, District of Columbia, and Rhode Island.

^*Significantly different than rate for children in states with no family coverage at the 0.05 level.

Our additional analyses indicate that these differences are attributable to family coverage differences and not to other unmeasured differences across states. First, we do not find analogous participation differentials across states when we examine TANF-related eligibles (Table 2). The TANF-related eligible children in states with no family coverage expansions are only about 4 percentage points less likely to participate in Medicaid compared with children in states with family coverage expansions. Moreover, the cross-state participation differentials among TANF-related eligibles are not statistically significant in either the univariate or multivariate analyses.

Table 2.

Medicaid Participation Rates for TANF- and Poverty-Related Eligible Children by Family Coverage Status, 1999

	No Family Coverage	Family Coverage State-Funded Program	Family Coverage Medicaid Program
Poverty-Related Eligibles	57.1	78.5	80.8
TANF-Related Eligibles	77.0	79.5	81.8
Difference across Family Coverage Types1
Poverty-related
Unadjusted		21.4a	23.6a
Adjusted2		23.6a	24.0a
TANF-related
Unadjusted		2.5	4.9
Adjusted		3.5	4.6
Difference between Poverty-Related and TANF-Related within Family Coverage Type3
Unadjusted	−19.8b	−1.0	−1.1
Adjusted	−8.8b	4.3	3.7
Difference across Family Coverage Types Net of Underlying Differences across States4
Unadjusted		18.9c	18.7c
Adjusted		20.0c	19.5c

^aRate is significantly different than rate for those in states with no family coverage at the 0.05 level.

^bRate for poverty-related eligibles is different than the rate for TANF-related eligibles in the same family coverage type at the 0.05 level.

^cRate for poverty-related eligibles in states with family coverage is significantly different than rates for those in TANF-related at the 0.05 level.

¹Represents participation rates for children in states with family coverage minus participation rates from children in states with no family coverage.

²Regression model controls for age, race, and health status of child; work status, education and nativity of parent; and income, number of children, and welfare history of family.

³Represents participation rate for poverty-related children minus participation rate for TANF-related children among states with the same family coverage type.

⁴Represents difference in participation rates between states with family coverage and states with no family coverage for poverty-related children minus same difference for TANF-related children.

Second, we find that TANF-related and poverty-related eligibles have comparable participation rates in states with family coverage policies whereas there are large, statistically significant participation differences between the two groups in states that do not have family coverage expansions. In the multivariate models, for example, there is about a 4 percentage point difference between the participation rates of TANF and poverty-related children in the states with family coverage that is not statistically significant. This contrasts with an 8.8 percentage point difference in states without family coverage expansions, which is statistically significant.

Finally, when we use multivariate methods to net out the underlying differences in participation across states with and without family coverage we find that participation rates for children eligible under the poverty-related expansions are approximately 20 percentage points higher in states that cover the parents of these children under either a state-funded program or the Medicaid program, than in states that do not cover the parents of these children. Thus, these cross-sectional analyses indicate that family coverage expansions do raise participation levels among already eligible children and that the estimated effect is substantial.

The Massachusetts Before and After Comparison

The Massachusetts experience confirms the finding from the comparison of state groups that expanding coverage to parents leads to greater participation among children. Table 3 compares changes in insurance coverage between 1997 and 1999 for children eligible for Medicaid under the poverty-related expansions in Massachusetts and in the rest of the nation. As can be seen, coverage under Medicaid rose from 42.3 percent to 63.6 percent of these eligible children in Massachusetts, a 21.3 percentage point increase. While Medicaid coverage also rose in the rest of the nation, the increase was a much smaller 3.6 percentage points. Private coverage for this group fell by 11.3 percentage points in Massachusetts compared to 4.4 percentage points in the rest of the nation. Finally, the uninsurance rate for children eligible for Medicaid under the poverty-related expansions fell by 9.7 percentage points in Massachusetts but remained nearly unchanged in the rest of the nation.

Table 3.

Changes in Insurance Distribution of Children Eligible under the Poverty-Related Expansions, Massachusetts and the Rest of the Nation, 1997–1999

					Difference-in-Difference
	Massachusetts		Rest of the Nation
	1997	1999	1997	1999	Unadjusted	Regression Adjusted with Offers1	Regression Adjusted without Offers2
Medicaid	42.3	63.6	30.3	33.9	17.7*	14.2*	15.4*
Uninsured	18.1	8.4	21.7	22.6	−10.6*	−11.0*	−9.3*
Private	39.3	28.0	48.0	43.6	−7.1	−3.2	−6.1
Crowd-Out Estimate						22.5%	39.6%
Estimate

¹Regressions include controls for year, year^*Massachusetts, Massachusetts, age, race, and health status of child; work status, education and nativity of parent; and income, number of children, welfare history of the family, and whether the parents have an offer of employer-sponsored coverage.

²Regressions include above mentioned controls but do not control for whether the parents have an offer of employer-sponsored coverage.

^*Changes in rate for Massachusetts is significantly different than changes in the rate for the rest of the nation at the 0.05 level.

Turning to the regression-adjusted difference-in-difference estimates, we find that Medicaid coverage in Massachusetts increased by 14.2 percentage points relative to the change in the rest of the nation. Massachusetts also saw a decline in the uninsurance rate for these children that was 11.0 percentage points greater than those observed elsewhere. Both of these differences are statistically significant. While the difference-in-difference estimates indicate that these eligible children lost private coverage at a somewhat higher rate in Massachusetts relative to the rest of the nation, the 3.2 percentage point differential was not statistically significant.

Most of the increased Medicaid participation was due to reductions in uninsurance rather than substitution of private coverage. The difference-in-difference results indicate a lower-bound estimate for the substitution of 22.5 percent of private coverage. When we do not control for offers of employer-sponsored coverage, we derive an upper-bound substitution effect of 39.6 percent. While the actual substitution rate was likely somewhere between these two estimates, as mentioned previously, the prior literature leads us to put more weight on the lower estimate.

POLICY IMPLICATIONS

The findings here make it clear that expanding the Medicaid program to include parents presents a very promising policy opportunity. By covering parents, states may achieve the dual goals of increasing Medicaid coverage of eligible but uninsured children and extending new coverage to parents. The share of poor parents who are uninsured has increased dramatically since welfare reform and more than 40 percent were uninsured in 2000 compared to 27 percent of poor children (Holahan, Dubay, and Kenney 2003). Expanding public coverage to parents up to the same income level at which their children are currently eligible would provide clear benefits to the 5.6 million and 1.8 million uninsured parents who would be made eligible for Medicaid and SCHIP, respectively (Dubay and Kenney 2001a). Furthermore, such expansions would reduce within-family inequities in coverage (Davidoff et al. 2001).

Expansions to parents are likely to provide benefits to children as well, because family coverage policies appear to lead to greater participation in Medicaid and lower uninsurance rates among eligible children. In 1999, for example, children eligible for Medicaid under the poverty-related expansions participated in Medicaid at a rate that was over 20 percentage points higher in states that expanded coverage to parents than in states that did not. In Massachusetts, implementation of family coverage led to a 14.2 percentage point increase in Medicaid coverage and an 11 percentage point reduction in uninsurance among children eligible for Medicaid under the poverty-related expansions, relative to the rest of the country. Reaching more of the uninsured children who are already eligible for Medicaid is critical to reducing uninsurance among children, since over half (52 percent) of all uninsured children are eligible for Medicaid (Dubay, Haley, and Kenney 2002).

Importantly, there is a possibility that expanding coverage will draw some children and parents in who were not uninsured but were paying for private coverage—the substitution problem so much discussed when program expansions are on the table. Policymakers need to come to terms with the inevitable fact that covering whole families may lead to some substitution of public for what was previously privately financed coverage. Results from Massachusetts suggest that the extent of substitution will be small—about the magnitude observed under the Medicaid expansions for children (Dubay 1999)—with most of the increased coverage coming from real reductions in the number of uninsured children.

Coverage expansions to parents with higher incomes than those examined here may be expected to increase the amount of substitution, while coverage expansions directed at lower income parents should result in lower levels of substitution. It is important to note, however, that for low-income families, substituting public for private insurance may yield large benefits—not only in reduced premiums, copayments, and deductions, but also in better benefits, including broader coverage of routine and other preventive care (Dubay and Kenney 2001b). Some substitution may be the price society has to pay to achieve health insurance for the nearly 40 million Americans currently uninsured.

One potential caveat to our analysis is that our findings come from a limited number of states that have been ahead of the curve in family coverage expansions and may have invested more heavily in outreach efforts and enrollment simplification than other states that have subsequently expanded coverage to parents. Several states have expanded eligibility to parents since 1999 and it will be important to assess whether the findings reported here repeat themselves. It will be also be important to evaluate the impact of these more recent expansions on uninsurance among parents and to assess the extent to which they resulted in improvements in access to care as well.