Insurance Coverage & Costs
Access to Care
Uninsured and Low-Income
Racial/Ethnic Disparities
Safety Net Providers
Community Health Centers
Hospitals
Physicians
Insured People
Quality & Care Delivery
Health Care Markets
Issue Briefs
Data Bulletins
Research Briefs
Policy Analyses
Community Reports
Journal Articles
Other Publications
Surveys
Site Visits
Design and Methods
Data Files
|
||
CHIPing Away at the Problem of Uninsured Children Issue Brief No. 14 |
||
Who Declines Employer-Sponsored Health Insurance and Is Uninsured? Issue Brief No. 22 |
||
Medicaid Eligibility Policy and the Crowding-Out Effect: Issue Brief No. 03 |
||
Some Communities Make Progress in Reducing Children's Uninsurance Data Bulletin No. 19 |
||
Tracking Recent Changes in Health Coverage for Low-Income Children with the Community Tracking Study, 1996-1997 and 1998-1999
Research Report No. 04
April 2000
Peter J. Cunningham, Michael H. Park
he Center for Studying Health System Change is supported in full by The Robert Wood
Johnson Foundation and is affiliated with Mathematica Policy Research, Inc. The authors
would like to thank the following people for providing comments on earlier drafts of this
report: Paul Ginsburg, Jim Reschovsky, Joy Grossman and Ann Greiner from HSC; Linda
Billheimer, Bob Hughes, Jim Knickman and Maureen Michael from The Robert Wood Johnson
Foundation; Genevieve Kenney, Steve Zuckerman and Lisa Dubay from the Urban Institute. Ase
Sewell, Jenny Zhang and Beny Wu of Social and Scientific Systems, Inc., provided excellent
programming assistance.
Contents
Explaining Changes in Coverage Among Low-Income Children
©2000 Center for Studying Health System Change
Whether SCHIP leads to reductions in the number of uninsured children depends not only on how successful the state programs are in enrolling eligible children, but also on whether gains in coverage through SCHIP and other expansions of public coverage are offset by other factors. For example, Medicaid coverage of poor children decreased during the mid-1990s, which some attribute to state and federal welfare reform efforts (Fronstin, 2000). Indeed, recent research has shown that uninsurance among former welfare recipients increases the longer they have been off welfare, primarily because many lose Medicaid coverage upon leaving welfare and are unable to obtain private insurance coverage through their employer or elsewhere (Garrett and Holahan, 2000).
In addition, gains in public coverage may not lead to lower uninsurance rates among low-income children if there are also decreases in private insurance coverage. These decreases may result either from substitution of public for private coverage (i.e., because of increased eligibility for public programs) or because of independent factors leading to lower private insurance coverage. Independent factors may include rising employee contributions to employer-sponsored coverage, changes in the labor market that affect access to employer-sponsored coverage or changes in underlying characteristics of children (e.g., income and racial/ethnic composition) that are associated with private insurance coverage.
An examination of recent trends in the health insurance coverage of low-income children is needed to determine whether there have been increases in public coverage (possibly due to SCHIP and Medicaid expansions), and whether these increases are resulting in decreases in the number of uninsured children or they are being offset by other factors. Data from the 1996-1997 and 1998-1999 Community Tracking Study (CTS) Household Surveys—which contain the most up-to-date estimates of children’s health insurance coverage—show that while the percentage of low-income children with public coverage has increased, the percentage with private insurance coverage has decreased, resulting in no net change in the percentage who are uninsured. CTS data from the two surveys are used to explore reasons for these changes in coverage among low-income children.
METHODS
Data Source
The CTS is a longitudinal study designed to track changes in the health care system and the effects of these changes on individuals. The CTS Household Survey is designed to produce representative estimates for the U.S. population and 60 randomly communities of individual’s health insurance coverage, access to care, use of services and perceived quality (Strouse et al., 1998). The sample for the surveys was obtained primarily through random digit dialing, supplemented by in-person interviews to represent households without telephones.
As of this writing, two rounds of the survey have been completed. The first round of the Household Survey took place between August 1996 and August 1997, while Round 2 took place between August 1998 and October 1999, approximately a two-year interval between rounds.
The timing of the two surveys also allows for an examination of how health insurance coverage of children has changed during the early stages of SCHIP implementation. The field period for Round 1 ended just before passage of the SCHIP legislation in 1997. The field period for Round 2 began after 33 states had begun implementing their SCHIP programs and were enrolling children. These include all of the largest states, such as New York, California, Illinois, Pennsylvania, Florida and Texas. In addition, all but five states had implemented their programs by January 1, 1999 (slightly less than the halfway point of the survey field period). Nevertheless, because some states are implementing their SCHIP programs gradually, and their full effects may not be seen for some time, these results likely reflect the early stages of SCHIP implementation.
Survey instruments and procedures used in both rounds of the survey were nearly identical, thus ensuring that any changes observed are not due to changes in survey methodology. Information was obtained on all adults in a family, as well as one randomly selected child. Information on children’s health insurance coverage was provided by their parents. The number of children in the sample includes about 10,600 in Round 1 and 10,300 in Round 2. The number of low-income children (i.e., in families with incomes below 200 percent of poverty) includes about 3,900 in Round 1 and 3,500 in Round 2. The overall response rate for both rounds of the survey was 65 percent. Estimates are weighted to reflect all U.S. children below the age of 18 and to account for survey nonresponses. Standard errors used in tests of statistical significance take into account the complex survey design, including the clustering of most of the sample into 60 sites, the national supplement, mixed sample frames (random digit dialing and in-person interviews) and the selection of multiple families within a household.
Classification of Health Insurance Coverage
In this report, health insurance coverage of children is estimated based on the day of the interview (i.e., a point-in-time estimate). Health insurance coverage is defined based on the following mutually exclusive categories: (1) employer-sponsored private insurance; (2) other forms of private insurance (e.g., direct purchase of nongroup coverage); (3) Medicaid and other state insurance programs (including SCHIP programs); (4) other forms of public coverage, including Medicare, military coverage (CHAMPUS) and the Indian Health Service (IHS); and (5) uninsured.
While the survey instrument allows for multiple sources of coverage to be reported, individuals with multiple sources of coverage were assigned to a single category based on the following order of priority: Medicare; employer-sponsored coverage; other private insurance; other public (military, IHS); and Medicaid and other state coverage. Thus, children were classified as "Medicaid and other state coverage" only if they had no other form of public or private health insurance coverage.
State-specific names for Medicaid and other state insurance programs (including SCHIP
programs) were used in the questions for public coverage. However, because many states
have used funds through SCHIP to expand Medicaid rather than to create separate programs,
it is difficult to distinguish in the survey between children covered through SCHIP or
through traditional Medicaid. Therefore, we present combined estimates of Medicaid and
other state coverage.
CHANGES IN COVERAGE
For children in families with incomes below 200 percent of the federal poverty level, public coverage increased significantly at the same time that private insurance coverage decreased, resulting in no significant change in the proportion of low-income children who were uninsured. The proportion of low-income children with Medicaid and other state coverage increased from 29 percent to 32.9 percent, while the proportion of low-income children with any private coverage decreased from 47 percent to 41.7 percent (Table 1; standard errors are shown in the Appendix). Most of the decrease in private insurance coverage was the result of decreases in employer-sponsored coverage (from 38.4 percent to 34.4 percent). The increase in public coverage and decrease in private insurance resulted in no significant change in the proportion of low-income children who were uninsured between 1996-1997 and 1998-1999 (between 19 percent and 20 percent in both surveys).
Moreover, the increase in public coverage and decrease in private coverage occurred primarily for children who were the primary targets for SCHIP—those in
Table 1: Health Insurance Coverage of Children
| Insurance Type | Percent 1996-1997 | Percent 1998-1999 |
| All children | ||
| Private insurance | 69.9 |
68.5** |
|
61.1 |
60.8 |
|
8.8 |
7.7* |
| Medicaid and other state coverage | 14.8 |
16.0** |
| Other public2 | 3.6 |
4.0 |
| Uninsured | 11.7 |
11.5 |
| Less than 200% of poverty | ||
| Private insurance | 47.0 |
41.7* |
|
38.4 |
34.4* |
|
8.6 |
7.4 |
| Medicaid and other state coverage | 29.0 |
32.9* |
| Other public2 | 4.6 |
5.6 |
| Uninsured | 19.4 |
19.8 |
| 200% of poverty and higher | ||
| Private insurance | 88.7 |
89.0 |
|
79.7 |
81.0 |
|
8.9 |
8.0** |
| Medicaid and other state coverage | 3.2 |
3.1 |
| Other public2 | 2.8 |
2.9 |
| Uninsured | 5.3 |
5.1 |
*Difference from 1996-1997 estimate is statistically significant at .05 level.
**Difference from 1996-1997 estimate is statistically significant at .10 level.
1
Includes nongroup private insurance and private insurance obtained through someone outside of the household.2Includes Medicare, CHAMPUS, Indian Health Service and any other unspecified coverage.
Source: Community Tracking Study Household Survey, 1996-1997 and 1998-1999 families with incomes between 100 percent and 199 percent of poverty.
For children below the poverty line, there was no change in the percentage with Medicaid and other state coverage (just under 50 percent in both surveys), while private insurance decreased. It is possible that any gains in public coverage through SCHIP and Medicaid expansions for poor children were offset by losses of Medicaid coverage due to welfare reform, although this could not be examined explicitly with the CTS data.
The decrease in private insurance coverage observed for low-income children does not hold for moderate- and high-income children (i.e., family incomes at 200 percent of poverty and higher), although there was a slight decrease in "other private coverage" among children in this income group (from 8.9 percent to 8 percent). Even children closest to being low-income (i.e., in families with incomes between 200 percent and 299 percent of poverty), had no change in their private and public coverage, or in their uninsurance rates. In sum, the most substantial changes in public and private coverage between 1996-1997 and 1998-1999 occurred mainly for children in families with incomes between 100 percent and 199 percent of the poverty level.
Changes in Coverage for Low-Income Parents
Private insurance coverage also decreased for the parents of low-income children, from 50.6 percent in 1996-1997 to 46 percent in 1998-99 (Table 2). However, because low-income parents have fewer public coverage alternatives than do their children, there was no significant change in their rates of public coverage. As a result, uninsurance rates among low-income parents increased between 1996-1997 and 1998-1999 (from 30.5 percent to 35.3 percent).
Table 2: Health Insurance Coverage of Adults (age 21-64)
| Percent 1996-1997 | Percent 1998-1999 | |
| All adults | ||
| Adults with children | ||
| Private insurance | 75.1 |
75.2 |
|
69.1 |
69.8 |
|
6.1 |
5.4** |
| Medicaid and other state coverage | 5.8 |
4.9* |
| Other public2 | 3.4 |
3.4 |
| Uninsured | 15.8 |
16.6 |
| Adults with no children | ||
| Private insurance | 71.5 |
72.7* |
Employer-sponsored |
62.9 |
64.4* |
Other private1 |
8.6 |
8.2 |
| Medicaid and other state coverage | 2.8 |
2.8 |
| Other public2 | 8.5 |
8.5 |
| Uninsured | 17.2 |
16.1* |
| Low income adults | ||
| Adults with children | ||
| Private insurance | 50.6 |
46.0* |
|
45.7 |
41.8* |
|
4.9 |
4.2 |
| Medicaid and other state coverage | 13.7 |
12.6 |
| Other public2 | 5.3 |
6.0 |
| Uninsured | 30.5 |
35.3* |
| Adults with no children | ||
| Private insurance | 38.9 |
38.5 |
|
29.6 |
29.3 |
|
9.3 |
9.2 |
| Medicaid and other state coverage | 8.9 |
9.3 |
| Other public2 | 17.2 |
17.8 |
| Uninsured | 35.0 |
34.4 |
*Difference from 1996-1997 estimate is statistically significant at .05 level.
**Difference from 1996-1997 estimate is statistically significant at .10 level.
1
Includes nongroup private insurance and private insurance obtained through someone outside of the household.2Includes Medicare, CHAMPUS, Indian Health Service and any other unspecified coverage.
Source: Community Tracking Study Household Survey, 1996-1997 and 1998-1999
Moreover, the decrease in private insurance and the increase in uninsurance rates were observed only for low-income parents with children. Although low-income adults without children have lower rates of private coverage than low-income adults with children, the sharp decreases in private insurance coverage observed for low-income adults with children were not observed for other low-income adults. In sum, sizable decreases in private insurance coverage between the two surveys were experienced almost exclusively by low-income families with children.
Changes in Access to and Enrollment in Employer-Sponsored Coverage among Low- Income Children
The decline in employer-sponsored coverage among low-income children may be the result of fewer low-income children having access to employer-sponsored coverage, or fewer children being enrolled in this coverage by their parents when it is available. Changes in access to employer-sponsored coverage may result from changes in either parents’ employment status or the relative number of employed parents who are offered and eligible for coverage through their employer.
To examine this, data in the survey on parents’ employment status and whether they are offered and eligible for coverage were linked to individual children. Thus, access to employer-sponsored coverage is defined as the percent of low-income children with at least one parent who was offered and eligible for coverage through an employer. By focusing on children with access to employer-sponsored coverage, we can observe whether there were changes between the two surveys in the percent of low-income children enrolled in coverage that was available to them (i.e., the take-up rate), as well as changes in the percent who opted for other private or public coverage, or who were uninsured.
The results show a small decrease between the two surveys in the proportion of all low-income children with access to employer-sponsored coverage (from 47.9 percent to 45.5 percent) (Table 3). This was mainly the result of fewer employed parents being offered coverage through an employer (from 70.6 percent in 1996-1997 to 66.1 percent in 1998-1999), rather than fewer parents being employed. In fact, the percent of low-income children with at least one parent employed full-time increased slightly, although this change was not statistically significant.
Table 3. Access to Employer-Sponsored Coverage among Low-Income Children
1996-1997 |
1998-1999 |
|
| Overall percent with access to employer-sponsored coverage1 | 47.9 |
45.5** |
| Percent with at least one parent employed full-time2 | 62.4 |
64.0 |
| Percent with employed parent who is offered and eligible for coverage3 | 70.6 |
66.1* |
*Difference from 1996-1997 estimate is statistically significant at .05 level.
**Difference from 1996-1997 is statistically significant at .10 level.
1
Sample includes all low-income children. Access to employer-sponsored coverage defined as having one or both parents offered and eligible for employer-sponsored coverage.2Employed full-time includes one or both parents working at least 35 hours per week.
3Sample includes low-income children (< 200 percent of poverty) who have at least one employed parent who is not self-employed.
Source: Community Tracking Study Household Survey, 1996-1997 and 1998-1999.
Even among low-income children with access to employer-sponsored coverage, fewer were being enrolled in that coverage and more of their parents were opting instead for Medicaid and other state coverage. Among low-income children whose parents were offered and eligible for employer-sponsored coverage, 72 percent were enrolled in that coverage in 1996-1997, compared with 66.2 percent in 1998-1999 (Table 4). Conversely, the proportion of low-income children with access to employer-sponsored coverage who were enrolled in Medicaid or other state coverage increased from 9.9 percent in 1996-1997 to 14.3 percent in 1998-1999. About 11 percent of children with access to
Table 4. Health Insurance Coverage of Low-Income Children (< 200 percent of poverty) with Access to Employer-Sponsored Coverage1
1996-1997 |
1998-1999 |
|
| Percent enrolled in coverage offered to parent | 72.0 |
66.2* |
| Percent enrolled in other private coverage | 4.8 |
5.5 |
| Percent enrolled in Medicaid or other state coverage | 9.9 |
14.3* |
| Percent enrolled in other public coverage | 2.7 |
3.1 |
| Percent uninsured | 10.6 |
10.8 |
*Difference from 1996-1997 estimate is statistically significant at .05 level.
1
Sample includes low-income children with at least one parent who is offered and eligible for employer-sponsored coverage.Source: Community Tracking Study Household Survey, 1996-1997 and 1998-1999
employer-sponsored coverage went without any insurance, with that number holding steady across both time periods.
Further analysis revealed that declining take-up rates account for about two-thirds of the decrease in employer-sponsored coverage among low-income children. If the take-up rate had remained unchanged between 1996-1997 and 1998-1999 (but access had declined), then the percent of low-income children with employer-sponsored coverage would have declined by only 1.4 percentage points (from 38.4 percent in 1996-1997 to 37 percent in 1998-1999), compared to the actual decline of about 4 percentage points. Conversely, if the proportion with access to employer-sponsored coverage had remained unchanged (but take-up rates had declined), then the percent with employer-sponsored coverage would have declined by 2.6 percentage points (to about 35.9 percent in 1998-1999).
These results may indicate the continuation of a longer-term trend in which falling take-up rates have been identified as a major factor in accounting for the decreases in employer-sponsored coverage for the general population between the late 1980s and mid-1990s (Cooper and Schone, 1997). The CTS data are also consistent with previous research in that decreases in take-up rates over time are occurring primarily for lower-income persons.
EXPLAINING CHANGES IN COVERAGE AMONG LOW-INCOME CHILDREN
The changes in public and private coverage observed for low-income children may be due to a number of different factors. In this section, we discuss several of these, including changes in key labor market factors—such as the size of the firm—that may affect availability of and enrollment in employer-sponsored coverage; changes in key population characteristics (e.g., income) that are associated with health insurance coverage; increases in private health insurance premiums; and substitution of public for private coverage. Evidence from the CTS surveys is also provided to assess the relative importance of some of these factors in explaining changes in coverage for low-income children.
Changes in the Size of Firm Where Low-Income Parents Are Employed
Decreases in access to and take-up of employer-sponsored coverage among low-income children do not necessarily reflect less willingness of employers of low-income parents to offer coverage, nor does it necessarily reflect less willingness among parents to enroll their children in this coverage when it is available. Rather, these decreases could also occur if there were shifts in the types of jobs in which parents of low-income children were employed.
Specifically, the size of the firm is one of the most important factors related to whether employers offer coverage (Cantor et al., 1995; Long and Marquis 1993). Smaller firms are less likely to offer coverage to their employees than are larger firms or the public sector, primarily because small employers pay higher premiums relative to the benefits received (Gabel et al., 1999). Employees in small firms also tend to pay a larger share of the premium than employees in larger firms, which no doubt contributes to the somewhat lower take-up rates among employees in small firms (Gabel et al., 1997; Cooper and Schone, 1997). Thus, decreases in access to and enrollment in employer-sponsored coverage could result from a shift in employment from large to small firms among the parents of low-income children, rather than less willingness among employers to offer coverage and less willingness among parents to accept it.
To examine this, data on the size of the firm for employed parents are linked to their children. Since both parents of the child may be working for employers that differ in type and size, children are classified hierarchically by firm size, beginning with firms that are most likely to offer health insurance coverage. Thus, children are classified in the largest firm size category if they have at least one parent employed in the public sector or in firms with 100 or more workers. All other children are then classified incrementally based on whether they have a parent employed in the next largest firm size, beginning with the 50-99 worker category. Children are classified as self-employed if they have working parents who are self-employed only.
The results show that there was a decrease in the proportion of low-income children with parents employed in the type of firms that are the most likely to offer coverage (government or private firms with 100 or more workers)—from 62.4 percent in 1996-1997 to 58.1 percent in 1998-1999 (Table 5). By contrast, there was an increase in the percentage of low-income children with parents employed in firms with 10-24 workers (8.3 percent to 10.1 percent) and 25-49 workers (6.1 percent to 8.3 percent).
However, most of the decrease in offer and take-up rates appears to be unrelated to changes in the size of the firm where low-income parents work. While small sample sizes do not allow for a more detailed analysis of changes in offer and take-up rates by firm size for low-income children, further analysis revealed that most of the decline in offer and take-up rates among low-income children would still have occurred even if there had been no changes in the size of the firm where low-income parents worked.
Table 5. Employment Characteristics of Low-Income Children’s Parents
Percent of low-income children |
||
1996-1997 |
1998-1999 |
|
| At least one parent employed in government or in a private firm with 100 or more workers | 62.4 |
58.1* |
| At least one parent employed in firm with 50-99 workers | 5.2 |
5.4 |
| At least one parent employed in firm with 25-49 workers | 6.1 |
8.3* |
| At least one parent employed in firm with 10-24 workers | 8.3 |
10.1* |
| At least one parent employed in firm with fewer than 10 workers | 9.5 |
9.8 |
| All working parents self-employed | 8.6 |
8.3 |
*Difference from 1996-1997 estimate is statistically significant at .05 level.
Sample: Low-income children with at least one employed parent.
Note: To take into account the fact that a child may have two working parents who are employed in different-size firms, the categories were constructed hierarchically. Thus, children were assigned to the first category if any parent was employed in government or in a firm with 100 or more workers. For all other children, a child was assigned to the 50-99 category if any parent was employed in a firm with 50-99 workers. Children were assigned to the other firm size categories in a similar manner. Children were assigned to the self-employed category only if all working parents were self-employed.
Source: Community Tracking Study Household Survey, 1996-1997 and 1998-1999
Changes in the Characteristics of Children
Changes in private and public coverage among low-income children may also occur if there were changes in key sociodemographic characteristics that are also associated with health insurance coverage. For example, the effects of welfare reform, and strong economic growth and declining unemployment in the period between the two surveys may have resulted in some shifting of children out of poverty or low-income status, or changed the income distribution of children within the poverty and other low-income categories. Any changes in the income distribution of low-income children may affect the relative number who are eligible for public programs or are able to obtain private insurance coverage.
Similarly, rates of private and public health insurance coverage vary by racial and ethnic group and the age of the child (independent of socioeconomic factors), and any changes in these factors between the two surveys could also result in a change in coverage for all low-income children.
Table 6 shows that there were only small changes between the two surveys on these three characteristics. There was a small shift of children from between 100 percent and 149 percent of poverty to more children in the higher-income category (150 percent and 199 percent of poverty), although such a shift would suggest greater access to private insurance coverage rather than less. There was also a small decrease in the percentage of white children and an increase in the percentage who were Hispanic, as well as a small decrease in the percentage of younger children (although the last change was not statistically significant).
Table 6. Characteristics of Low-Income Children
| Percent 1996-1997 | Percent 1998- 1999 | |
| Family income as a percent of the federal poverty level | ||
| < 50% | 19.2 |
19.4 |
| 50-99% | 24.5 |
24.8 |
| 100-149% | 28.8 |
26.2* |
| 150-199% | 27.5 |
29.6** |
| Race/ethnicity | ||
| White | 52.6 |
49.6* |
| Black | 21.4 |
22.1 |
| Hispanic | 21.1 |
24.3* |
| Other | 4.9 |
4.1 |
| Age | ||
| 0-5 | 34.5 |
32.5 |
| 6-17 | 59.0 |
61.2 |
*Difference from 1996-1997 estimate is statistically significant at .05 level.
**Difference from 1996-1997 is statistically significant at .10 level.
Source: Community Tracking Study Household Survey, 1996-1997 and 1998-1999
To examine whether these small changes in the characteristics of low-income children accounted for any of the changes in rates of health insurance coverage, we simulated health insurance coverage for low-income children in 1998-1999, assuming they had the same characteristics for children observed in 1996-1997. In other words, would we have observed the same changes in private and public health coverage even if there had been no changes in income, race/ethnicity and age? The simulated rates were virtually identical to the actual rates of coverage in 1998-1999, indicating that the small changes on these three characteristics had virtually no effect on changes in coverage for low-income children.
Changes in Private Health Insurance Premiums
Decreases in private insurance coverage among low-income children may be due in part to increases in health insurance costs, which many consider to be the primary reason for the longer-term decline in employer-sponsored coverage during the 1980s and early 1990s (Kronick and Gilmer, 1999). Although premiums were relatively stable during the mid-1990s, premiums for employer-sponsored insurance rose 3.3 percent in 1998 and 4.8 percent in 1999 (Ginsburg, 1999; KFF/HRET, 1999). Increases in premiums among small employers were even higher (5.2 percent in 1998 and 6.9 percent in 1999). Moreover, the employee’s share of the premium has been increasing for family coverage in recent years—from an average of $122 per month in 1996 to $145 in 1999—while the employee share for single coverage decreased slightly, from an average of $37 per month in 1996 to $35 per month in 1999. Although the impact of these recent premium increases on rates of private coverage cannot be explicitly tested with CTS data, they may explain in part why private coverage has decreased for low-income families with children but not for low-income adults with no children (who are more likely to have single coverage).
Substitution of Public for Private Coverage
Another possible explanation for the decrease in private insurance coverage and the increase in public coverage among low-income children is that expa nded eligibility for public coverage through SCHIP and Medicaid expansions has resulted in substitution of public coverage for private. The Medicaid expansions of the 1980s led to some substitution of public for private coverage, although the extent of this substitution appears to have been fairly small (Dubay, 1999). Nevertheless, the more recent expansions in public coverage may increase the potential for substitution, since these expansions are targeting a higher-income group that generally has greater access to private insurance coverage than those who were the primary targets of public expansions during the 1980s.
Substitution can manifest itself in a number of different ways. Wider availability of public coverage may influence employers to drop coverage altogether, increase the employee share of the premium or change benefits in a way that makes the offered plans less attractive to workers. The presence of new public programs may also induce eligible, lower-income families to forgo costly private coverage and choose public coverage for their children when it becomes available. Finally, such programs may also influence some low-income workers to seek higher-paying jobs without health benefits, so they can trade private coverage for higher wages and the promise of subsidized or free public health insurance coverage for their children.
If there is greater substitution of public coverage for private—regardless of how it occurs—one might expect to see an increase between the two surveys in the percentage of low-income children switching from private to public coverage. This can be examined directly in the CTS Household Survey since respondents were asked whether they were enrolled in their current coverage for the entire 12 months prior to the interview and, if not, what type of coverage they had immediately prior to their current coverage.
An analysis of these data show that, of low-income children enrolled in Medicaid and other state coverage, only about 2 percent had switched into that coverage directly from private insurance at some time during the year prior to the interview. More important, the rate of switching from private to public coverage did not change between the two surveys. In both surveys, almost all of the low-income children covered by Medicaid or other state coverage were either enrolled in that coverage for the entire year preceding the interview (85 percent) or they were uninsured immediately prior to enrolling in Medicaid or other state coverage (13 percent).
Still, movement from private to public coverage may be more complex and may play out over an extended period of time during which children may be uninsured for a brief period. This longer time frame may occur, in part, because states are required by the SCHIP legislation to prevent or reduce the potential for substitution through such measures as mandatory waiting periods. On the other hand, because the time between SCHIP implementation and the survey interviews was generally less than a year, it is unlikely that substitution taking place over a more prolonged period would be reflected in these results. More lead time would be required, especially if substitution were to occur through changes in employer health plan offerings or by low-income parents switching to jobs with no health benefits to get higher wages. If the increases in public coverage and decreases in private coverage are the result of substitution, it is likely that this substitution would reflect expansions in public coverage that occurred prior to SCHIP.
Other Factors
Other factors may account for some of the changes in coverage for low-income children, although these factors were not explicitly examined. For example, there may have been changes in other employment characteristics of low-income parents that are also strongly correlated with the firm’s decision to offer coverage, such as the typical wage level of the firm and industry (Gabel et al., 1999).
Similarly, other changes in the characteristics of low-income children not reflected by changes in average income, race/ethnicity or age may also account in part for the increase in public coverage. For example, increases in public coverage might be expected if state and federal welfare reform efforts and changes in the economy resulted in former welfare recipients making up a higher percentage of the low-income population in 1998-1999 than in 1996-1997 (even if the income distribution per se did not change). Former welfare recipients may be more knowledgeable about public programs and feel less stigmatized by enrolling in them than other low-income persons.
The size of the increase in public coverage among low-income children may also be somewhat surprising if one expects to see a greater lag between the time when SCHIP programs are implemented and when sizable numbers of newly eligible children begin to enroll. Therefore, it is also possible that the increase in public coverage is due in part to higher rates of participation in public programs among eligible children, rather than increased eligibility for these programs. Outreach efforts are increasingly important in the Medicaid and SCHIP programs due to concerns that many eligible uninsured children will not enroll (Selden et al., 1999).
CONCLUSION
Private insurance coverage among low-income children decreased during the late 1990s at the same time that public coverage increased, resulting in no net change in uninsurance rates among low-income children. Moreover, the decrease in private insurance among low-income children was not part of a general trend among all children or the U.S. population in general. Rather, decreases in private insurance coverage appear to have occurred primarily among low-income families with children.
Uninsurance rates among low-income children remained steady during the period between the two surveys despite targeted efforts to expand eligibility for coverage (mainly through SCHIP). However, these findings reflect only the early stages of SCHIP implementation, and it is possible that uninsurance rates will decline in the future as SCHIP enrollment continues.
Perhaps the key policy question is whether the decrease in private insurance coverage among low-income families was a result of expanded eligibility for public programs (i.e., substituting public for private coverage), or whether the decrease in private insurance coverage occurred independently of public coverage expansions. If the decrease in private insurance was largely due to substitution of public for private insurance, it would suggest that public coverage expansions have primarily benefited children who already had private insurance by providing them with a lower-cost alternative. This explanation would also imply that the cost required to reduce the number of uninsured children through these programs is considerably higher than expected.
However, if the decrease in private insurance coverage occurred independently of public coverage expansions (e.g., as a result of health insurance premium increases), it would imply that expansions in public coverage have provided an important safety net to many low-income children whose parents can no longer afford private insurance coverage. In other words, there would have been a substantial increase in uninsurance rates among low-income children (as has happened with their parents) if not for these expansions in public coverage.
The findings presented in this study do not allow us to determine conclusively which of these two alternative interpretations is more appropriate. Some amount of substitution of public for private coverage is to be expected in any type of incremental health insurance expansion that targets the uninsured. On the other hand, since these findings reflect the early stages of SCHIP implementation, it may be too early for a substantial amount of substitution to have taken place, especially given the fact that the vast majority of new enrollees in Medicaid and other state coverage were previously uninsured.
In addition, there are certainly pressures on low-income families to discontinue private insurance coverage that are independent of public coverage expansions, and these pressures are likely to increase in the future. Paying for health insurance—even if it is partially subsidized by employers—is a major financial burden for many low-income families. Other HSC research has shown that low-income persons are five to ten times more likely to decline employer-sponsored coverage than higher-income persons, and lower-wage workers often have to pay more for health insurance than higher-wage workers (Cunningham et al., 1999).
Further increases in health insurance costs—as many predict—will only increase the financial burden and, therefore, the pressure on low-income families to find alternatives to private insurance or do without insurance entirely. This will make it even more difficult to determine in the future whether the primary effect of public coverage expansions is to draw low-income persons away from private coverage, or whether these programs act as a safety net for families who can no longer afford private insurance.
REFERENCES
Cantor, J., S. Long and M.S. Marquis. 1995. "Private Employment-Based Health Insurance in Ten States." Health Affairs 14 (Summer): 199-211.
Cooper, P.F., and B.S. Schone. 1997. "More Offers, Fewer Takers for Employment-Based Health Insurance: 1987 and 1996." Health Affairs 16 (6): 142-49.
Cunningham, P., L. Schaefer and C. Hogan. 1999. "Who Declines Employer-Sponsored Health Insurance and Is Uninsured?" Issue Brief No. 22. Center for Studying Health System Change. Washington, D.C.
Dubay, L. 1999. "Expansions in Public Health Insurance and Crowd-out: What the Evidence Says." The Kaiser Project in Incremental Health Reform. Washington, D.C.: The Henry J. Kaiser Family Foundation:
Fronstin, P. 2000. "Sources of Health Insurance and Characteristics of the Uninsured."
EBRI Issue Brief #217. Washington D.C.: Employee Benefit Research Institute.
Gabel, J., K. Hurst, H. Whitmore and C. Hoffman. 1999. "Class and Benefits at the Workplace." Health Affairs 18 (3): 144-50.
Gabel, J., P.B. Ginsburg and K.A. Hunt. 1997. "Small Employers And Their Health Benefits, 1988-1996: An Awkward Adolescence." Health Affairs 16(5): 103-10.
Garrett, B., and J. Holahan. 2000. "Health Insurance Coverage After Welfare." Health Affairs 19 (1): 175-84.
Ginsburg, P. 1999. "Tracking Health Care Costs: Long-Predicted Upturn Appears."
Issue Brief #23. Center for Studying Health System Change. Washington, D.C.
Kaiser Family Foundation and Health Research and Education Trust. 1999. Employer Health Benefits: 1999 Annual Survey. Menlo Park, Calif., and Chicago, Ill.
Kronick, R., and T. Gilmer. 1999. "Explaining the Decline in Health Insurance Coverage, 1979-1995." Health Affairs 18(2): 30-47.
Long, S., and M.S. Marquis. 1993. "Gaps in Employer Coverage: Lack of Supply or Lack of Demand?" Health Affairs 12 (Supplement): 282-93.
Rosenbach, M., and K. Lewis. 1998. Estimates of Health Insurance Coverage in the Community Tracking Study and the Current Population Survey. Technical Publication #16. Center for Studying Health System Change. Washington, D.C.
Selden, T.M., J.S. Banthin and J.W. Cohen. 1999. "Waiting in the Wings: Eligibility and Enrollment in The State Children’s Health Insurance Program." Health Affairs 18(2): 126-33.
Strouse, R., J. Hall, F. Potter, B.L. Carlson, P. Cunningham, J. Pascale and R. Stapulonis. 1998. Report on Survey Methods for the Community Tracking Study’s 1996-1997 Round One Household Survey. Technical Publication No. 15 (November). Center for Studying Health System Change. Washington, D.C.
Appendix
Standard Errors of the Estimates for Children’s Health Insurance Coverage
(for percentages from Table 1)
| Insurance Type | Standard errors for 1996-1997 estimate |
Standard errors for 1998-1999 estimate |
Standard errors |
| All children | |||
| Private insurance | 1.1 |
1.0 |
0.8 |
|
1.1 |
1.1 |
0.8 |
|
0.4 |
0.3 |
0.4 |
| Medicaid and other state coverage | 0.8 |
0.8 |
0.6 |
| Other public3 | 0.3 |
0.5 |
0.3 |
| Uninsured | 0.6 |
0.6 |
0.5 |
Less than 200% of poverty |
|||
Private insurance |
1.4 |
1.2 |
1.6 |
|
1.2 |
1.3 |
1.4 |
|
0.5 |
0.5 |
0.8 |
| Medicaid and other state coverage | 1.5 |
1.2 |
1.3 |
| Other public2 | 0.5 |
0.9 |
0.8 |
| Uninsured | 1.1 |
1.2 |
1.1 |
200% of poverty and higher |
|||
Private insurance |
0.6 |
0.6 |
0.7 |
|
0.9 |
0.7 |
0.8 |
|
0.5 |
0.3 |
0.5 |
| Medicaid and other state coverage | 0.3 |
0.3 |
0.4 |
| Other public2 | 0.3 |
0.3 |
0.3 |
| Uninsured | 0.4 |
0.4 |
0.5 |
1Tests of change between 1996-1997 and 1998-1999 are based on the standard error
of the difference between 1996-1997 and 1998-1999 estimates and take into account the
individual correlation in the longitudinal component of the sample (see endnote 1).
2Includes nongroup private insurance and private insurance obtained through
someone outside of the household.
3Includes Medicare, CHAMPUS, Indian Health Service and any other unspecified
coverage.
Source: Community Tracking Study Household Survey, 1996-1997 and 1998-1999
ENDNOTES
1The Round 2 sample includes a longitudinal cohort from Round 1 (about 40 percent of the Round 2 sample) as well as a new cross-sectional sample, although the combined sample for Round 2 is designed to be representative of the population for 1998-1999. All estimates in this report from the two surveys are computed as cross-sectional estimates, and are weighted to represent the population for each time period. Tests of changes in coverage between the two surveys take into account the individual correlation in the longitudinal component of the sample. This reduces the variance of differences in means over time, which results in higher precision in the change estimates than if the two samples had been entirely independent.
2 Estimates of children’s health insurance coverage based on the CTS household survey differ somewhat from those of other surveys, such as the Current Population Survey (CPS). However, estimates from the two surveys are not directly comparable due to differences in the way that information about insurance is asked for and how it is reported. These differences include (1) the sequence of questions used to ask about health insurance coverage; (2) reference periods for reporting insurance information (CTS is based on coverage on the day of the interview, while coverage in CPS is reported for the entire year); (3) the timing of the survey field periods; and (4) classification of different types of health coverage. A more detailed analysis of the reasons for differences in estimates of health insurance coverage between CPS and the 1996-1997 survey is available in an HSC technical report (Rosenbach and Lewis, 1998).
3 There was also a slight increase in uninsurance rates for poor children between the two surveys, although this increase was not statistically significant due primarily to smaller sample sizes of poor children.
4 This was done by recomputing the offer and take-up rates for low-income children for 1998-1999 using the distribution of firm size for Round 1 and the offer and take-up rates (by firm size) for Round 2. The simulated results for Round 2 showed little difference from the actual results for Round 2.
5Specifically, separate regression models based on the 1998-1999 data were estimated for each category of health coverage (defined as binary dependent variables), and the three variables in Table 6 were included as categorical independent variables. Predictions for each of the dependent variables were derived by using the means of the independent variables for 1996-1997.
