Health Affairs, 22, no. 5 (2003): 198-209 doi: 10.1377/hlthaff.22.5.198 © 2003 by Project HOPE	New Online   Senate Health Reform Bill   Rewarding Providers   Public Option Policy Brief   Health Reform & Abortion   Delivery System Reform

This Article Abstract Reprint (PDF) Submit a response to this article Alert me when this article is cited Alert me when Comments are posted Alert me if a correction is posted Services E-mail this article to a friend Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Add to My Personal Archive Download to Citation Manager Reprints & Permissions Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Marquis, M. S. Articles by Kapur, K. Search for Related Content PubMed PubMed Citation Articles by Marquis, M. S. Articles by Kapur, K. Related Collections Insurance Coverage Insurance Coverage - Children Insurance - Employer-Based System Medicaid

Datawatch

Employment Transitions And Continuity of Health Insurance: Implications For Premium Assistance Programs

Abstract

 
We use data from two nationwide panel surveys to explore whether premium assistance programs can provide stable insurance for low-income children. We estimate that low-income children who are newly enrolled in an employer-group plan would keep that coverage longer than similar children keep newly acquired public insurance. We conclude that group coverage could provide a source of insurance for eligible low-income children that is more stable than public insurance. However, only one-third of low-income uninsured children have access to group insurance, and most low-income children with access to a group plan are enrolled in it. Thus, premium assistance programs are difficult to target effectively, and other programs are necessary to reach the majority of uninsured children.

A critical fact in designing programs to use the employment system to expand health insurance coverage is that employment is not static. When insurance is tied to employment, turnover in employment results in turnover in insurance. However, recent evidence suggests that continuity in coverage and care are just as important as having coverage is in improving access to health care.⁵ Linking coverage to the workplace will not promote continuity in coverage for some population groups. Job losses are more common among younger, lower-income, and less educated workers. In addition, most new jobs are short-term; about 20 percent last less than three months, and about 50 percent last less than a year.⁶ This turnover raises the question of when and for whom an employer insurance buy-in is consistent with the policy goal of continuity. Employment turnover also raises states’ cost of administering employer coverage buy-in programs.⁷

This paper examines the potential for premium assistance programs to insure low-income children. Although only a small share of the uninsured are eligible, as we demonstrate, such programs could help extend the number of uninsured people that a given public budget can assist. We investigate employment transitions among parents of low-income children who are eligible for group health insurance to predict the likely stability of this coverage. We also examine employees’ and employers’ characteristics that are related to employment transitions, to identify groups with stable employment and insurance. Since Medicaid is the primary alternative to premium assistance, we compare the duration of employer coverage to expected duration of Medicaid coverage for children who would be eligible for premium assistance. This provides a benchmark for assessing the expected continuity of such programs. Our analysis can help federal and state policymakers to establish eligibility criteria for premium assistance programs that would promote continuity of coverage and decrease the costs of administration.

Data And Methods

Top Data And Methods Methods Study Results Discussion And Policy... Editor's Notes NOTES

 
Data sources. The data for our study come from two nationwide panel studies of households: the 1996 Survey of Income and Program Participation (SIPP) and three panels of the Medical Expenditure Panel Survey (MEPS). The SIPP sample is a multistage stratified sample of the U.S. civilian noninstitutionalized population. The 1996 SIPP is a four-year panel of about 37,000 households. Households were interviewed twelve times from April 1996 through March 2000. Our analysis sample includes all children and their parents who completed the entire four-year panel.

For most analyses we focus on low-income children whose parents were eligible for employer-sponsored coverage at the start of the panel. However, the SIPP data do not definitively indicate jobs in which the parent was eligible for insurance if the parent did not enroll. We imputed eligibility for parents who were not in their group plan, based on the employee’s hours, business size, and industry. Those working fewer than twenty hours per week were assumed to be ineligible. We assigned offer rates to remaining uncovered workers using probabilities of offer by firm size and industry for uncovered employees estimated from the fifth interview of the SIPP panel. Our estimates of the share of low-income children with access to group coverage are consistent with the estimates from MEPS, based on reports of offer, as shown in the results. This gives us confidence that the error in classification from our imputation is relatively small.

Low-income families are those with incomes below 200 percent of the federal poverty level, a common threshold in SCHIP programs. Income is measured at the beginning of the panel to identify those who would be eligible for premium assistance at a point in time; we study subsequent transitions for this population. Our SIPP sample included 2,793 low-income children with access to employer-sponsored insurance, 881 of whom were not enrolled in the group plan.

MEPS uses an overlapping panel design, in which the sample selected in any given year is followed for two calendar years.⁸ For our study, we included the panels initiated in 1996, 1997, and 1998. Each family participated in five rounds of data collection. During each round surveyors collected information on all family members’ employment and health insurance coverage. MEPS includes a worker-specific question about employer coverage, which reflects the offer of insurance from the employer and the worker’s eligibility.⁹ MEPS does not ask whether family members are eligible for coverage in the plan; however, almost all employees—including low-wage employees—who are eligible for coverage are offered the option of family coverage.¹⁰ We restricted the MEPS sample to children and their parents who completed the two-year panel: 2,261 low-income children eligible for employer-sponsored insurance, of whom 1,052 were not enrolled in the plan.

Methods

Top Data And Methods Methods Study Results Discussion And Policy... Editor's Notes NOTES

 
We measured the remaining duration of jobs in progress at the initial interview for the primary parent (the parent providing employer-sponsored coverage or the parent working for the largest employer offering coverage). We take this as a measure of the expected length of coverage in the employer plan if the child was enrolled into the plan under a premium assistance program on the date of the interview. Because jobs with insurance coverage are known to last longer than jobs without it, we present job duration separately for families in which the parent and child are both insured by the employer, the parent only is insured, and neither parent nor child is insured.¹¹ The observed differences could reflect differences in the nature of the jobs or the characteristics of the parents in these groups, not just insurance status. To control for these other factors, we fit a multivariate logit model of hazard probabilities for discrete six-month time intervals.¹² Hazard probabilities are the likelihood of leaving the job, conditional on still having the job at the time. Each parent’s job-survival history is cast in six-month periods indicating whether the job ended in the period; the parent has a record for each period in which he or she held the job at the start of the period.¹³ For the SIPP analysis we have up to seven periods for each person; for the MEPS analysis we have up to three periods.¹⁴ We fit the logit model for job duration at the child level for the parents of low-income children with access to group coverage. The explanatory variables included the parent’s wage rate and indicators for the time period in question; parent or child, or both, covered by the group plan; family income above or below poverty; a new job (held less than one year at the initial interview); firm size; union job; industry; age of the child; and the occupation, education, age, and education of the parent. All of the explanatory variables are measured at the beginning of the panel to investigate how characteristics at the time eligibility is determined affect continuity. We also included interactions of the time-period indicators and new job, and the time periods and firm size.¹⁵ Marginal effects are shown as the average, predicted duration for all children in the estimation sample at each value of the characteristic.

To provide context for evaluating the implication of job duration for premium assistance programs, we investigated duration of Medicaid episodes for children who would be eligible for premium assistance. For uninsured children, the primary alternative to such assistance is expanded efforts to enroll the child in a public insurance program. Whether a premium assistance program promotes or impedes continuity of coverage for those eligible for premium assistance is relative to their expected continuity if enrolled in public programs. We measured the duration of all new Medicaid episodes (episodes that started after the first interview) among children whose parents were eligible for group coverage at the start of the episode.¹⁶ We included new episodes in all public insurance programs in our analysis of Medicaid episodes.¹⁷

In the SIPP data there were 768 new Medicaid episodes, and 443 in the MEPS data. We fit discrete-time hazard models for the duration of new Medicaid episodes using the explanatory variables described above, to be able to predict Medicaid duration for a population with the characteristics of the full population of children eligible for premium assistance.¹⁸ We use predicted durations from the fitted model, because differences in the characteristics of children starting new Medicaid episodes and all children eligible for premium assistance might bias conclusions based only on observed values. We restricted our prediction sample to those without group coverage, since these are the children that the premium assistance programs are intended to assist.

The opportunity to cover all family members in the same private plan might induce some families to move a child enrolled in a public program to the private (employer) plan. In such cases, we want to compare the expected remaining duration in the public plan with the duration of expected coverage in the employer plan. Therefore, we examined the remaining duration of Medicaid episodes in progress at the time of the initial interview among children with access to group coverage. There were 595 such episodes in the SIPP data and 447 in the MEPS data. Again, we fit discrete-time hazard models and adjust the estimates of duration of Medicaid episodes in progress to a population with the mix of characteristics that we observe for low-income children who are eligible for premium assistance.¹⁹ All differences described are statistically significant.

Study Results

Top Data And Methods Methods Study Results Discussion And Policy... Editor's Notes NOTES

 
Insurance status of children. About 14 percent of U.S. children were uninsured during 1996–1998, when our study sample members were initially interviewed (Exhibit 1). About two-thirds were in families with incomes below 200 percent of poverty; more than 20 percent of low-income children were uninsured. SCHIP has raised coverage rates among children since then. Nonetheless, about 27 percent of children in families below poverty and 17 percent of children in families with incomes at 100–200 percent of poverty are uninsured.²⁰

Employment duration: parents of low-income children eligible for employer coverage. Up to one-quarter of parents of low-income children will leave their current job within the year, and 25–30 percent will leave within eighteen months (Exhibit 2). We interpret this measure as the expected duration of coverage in a premium assistance program for a typical eligible child. Turnover is more rapid among parents in jobs that offer insurance if neither the parent nor the child is enrolled in the employer’s plan, with almost half leaving within eighteen months. In general, the two data sources provide a similar picture of duration. However, in the MEPS data, parents remain in insured jobs longer if the child is also enrolled in the plan than if the child is not covered—a difference not found in the SIPP data. About two-thirds of children whose parent leaves a job within a year would have access to new group coverage within a month, and about 80 percent would have access within four months (result not shown).²¹

Expected remaining job duration is much shorter for new jobs (those that had been held less than a year at the start of the panel) than for longer-held jobs. Exhibit 3 shows the expected or predicted remaining job duration for new jobs and older jobs at the beginning of the panel for a population with the demographic and job characteristics of the parents of low-income children with access to employer coverage who are not in the group plan—the group that is the target of premium assistance programs. About 25 percent of parents of this population of children are in new jobs. As many as half of such jobs end within a year, and enrolling chil dren through premium assistance plans would result in continuous coverage for less than a year for this proportion of newly enrolled children. However, only half as many longer-tenure jobs end within a year.

Medicaid duration. Our two data sources provide different pictures of the duration of Medicaid episodes; the SIPP data evidence shorter durations for both new and continuing episodes than the MEPS data do.²³ Nonetheless, comparing job duration with the length of Medicaid episodes in both databases suggests that premium assistance programs could provide a more stable source of coverage than public programs for children who have access to group coverage. Based on the MEPS data, only about 30 percent of eligible children would have expected coverage of less than a year if enrolled today in the parent’s group plan, but about 42 percent would be expected to leave a public insurance program within a year if newly enrolled today (Exhibits 3 and 4). Results in the SIPP data are even more dramatic. Duration in public programs is longer for children below poverty than for near-poor children. However, even for those below poverty, expected duration of coverage if enrolled in the parent’s group plan today exceeds that of new Medicaid episodes.

Discussion And Policy Implications

Top Data And Methods Methods Study Results Discussion And Policy... Editor's Notes NOTES

 
Providing low-income families with assistance to pay their share of the cost of employer coverage is intended to reduce uninsurance among children and promote family health insurance coverage. About one-third of uninsured low-income children have access to employer-sponsored coverage. While premium assistance programs offer a possible solution for only a minority of uninsured children, they remain attractive to many policymakers because they provide a way to leverage private dollars and extend the reach of public dollars. On the other hand, more than half of low-income children who have access to employer coverage are enrolled. Premium assistance for all low-income children who are eligible for group coverage, therefore, is not target-efficient; it may benefit many who are already insured. Restrictions on eligibility based on insurance status would add to administrative costs and raise issues of equity. Many would deem it inequitable to treat people in equal economic circumstances differently.

Turnover in employment would affect continuity of coverage under premium assistance programs. However, we found that only one-fourth of children would be expected to have coverage for less than a year, and most of those would have access to a new group plan within four months. For this latter group, coverage may be maintained, although job changes could cause some disruption in the type of coverage and providers. Moreover, administrative factors in qualifying a new employer plan for premium assistance could affect continuity of coverage even for those who move directly to a new job offering insurance.

Many believe that by providing health insurance on the job, premium assistance programs would promote attachment to the labor force among low-income parents and reduce turnover.²⁴ Turnover is less frequent among low-income parents with jobs in which both they and their children are enrolled in the employer-group plan than in other jobs eligible for group coverage, which lends some support to this hypothesis. However, when we control for other characteristics, differences in duration associated with insurance status of parent and child diminish. This suggests that there are factors other than nonparticipation in a group plan that lead to greater turnover, and these would persist even with premium assistance.

Our results suggest that program assistance may lead to more stable coverage than enrollment in public programs for some children. However, using job duration as a measure of continuity in premium assistance programs assumes that the parent will maintain coverage for the child, even if subsequent income changes make the family ineligible for premium assistance. The shorter duration in public programs could be attributable to income changes that could also affect eligibility for premium assistance. We have not developed a full model to explain turnover in public programs and its relationship to income fluctuations and job duration.²⁵ Further research on turnover in public programs as well as eligibility changes for premium assistance under different design options is necessary to identify the mix of private and public options that would be most effective in expanding coverage and promoting continuity of coverage for different subgroups. For example, rules related to fluctuating incomes to smooth insurance transitions might promote continuity in both public and public-private partnership programs.

In addition, we focused only on differences in continuity of coverage between public and private options and not on differences in access to care. Access depends in part on the generosity of benefits. Employer coverage could be less effective than public coverage in improving access for some if the employer coverage includes large deductibles or a less generous scope of benefits.

Some states have used income as a criterion in eligibility for premium assistance. Our results suggest that such criteria are not a means for improving continuity of care for the poorest children. Near-poor children are much more likely than the poorest children are to have access to employer coverage (59 percent versus 13 percent in SIPP, 50 percent versus 21 percent in MEPS), but expected continuity of coverage in group plans does not vary with income. However, our results are restricted to children with family incomes below 200 percent of poverty. There could be differences at higher income levels that would support income-related criteria to promote continuity. For example, the Maryland employer buy-in program is open to children with family incomes at 200–300 percent of poverty.²⁶

States wishing to limit the administrative cost of premium assistance programs associated with employment and coverage verification for workers who are most mobile might wish to consider waiting periods for new jobs before qualifying for premium assistance and direct the children of such workers to public programs. Job turnover is much higher for jobs that have been held less than a year than for jobs of longer duration. However, even in such cases, our results indicated that the expected duration of coverage in a new employer-group plan would equal or exceed the expected duration of a new episode of public insurance (Exhibits 3 and 4). In both the SIPP and the MEPS data, however, shifting a low-income child from a public coverage episode to group coverage in a newer job would decrease continuity. In addition, there are operational issues about waiting periods for premium assistance that would need to be explored, including the potential for later disruption in care and for encountering exclusions for preexisting conditions. Moreover, waiting periods could decrease continuity if uninsured children were directed to public programs and then switched to private programs at the end of a waiting period.

To date there has been limited real experience with employer buy-in programs. Our analysis suggests that these programs have the potential to reach about a third of uninsured low-income children, promote values of family coverage and continuity of coverage, and extend public money by leveraging the private system. However, as we note, effective targeting could lead to complex implementation issues. Participation in programs in states that have adopted these programs has been quite low, perhaps because of these difficulties.²⁷ We have focused only on premium assistance as a means to expand take-up among families eligible for group coverage. Some areas are developing premium assistance programs as a way of trying to encourage more employers of low-income workers to make coverage available; further research is needed to address this approach. With the flexibility offered by HIFA, more states might introduce premium assistance programs, providing the real experience needed to better understand the issues and the potential of these programs to reduce the numbers of uninsured American children.

Editor's Notes

Top Data And Methods Methods Study Results Discussion And Policy... Editor's Notes NOTES

 
Susan Marquis is a senior economist at RAND in Arlington, Virginia. Kanika Kapur is an associate economist at RAND in Santa Monica, California.

This study was supported by Contract no. J-9-P-7-0045 from the Pension and Welfare Benefits Administration, U.S. Department of Labor (DOL). Any views expressed herein are solely those of the authors, and no endorsement by the DOL or RAND is intended or should be inferred. The authors are grateful to Roald Euller for able assistance in constructing the SIPP data files used in this analysis.

NOTES

Top Data And Methods Methods Study Results Discussion And Policy... Editor's Notes NOTES

 

1. See, for example, Institute of Medicine, Health Insurance Is a Family Matter (Washington: National Academies Press, 2002); and K.L.Hanson, "Is Insurance for Children Enough? The Link between Parents’ and Children’s Health Care Revisited," Inquiry (Fall 1998): 294–302.
2. L.Ku and M. Broaddus, The Importance of Family-Based Insurance Expansions: New Research Findings about State Health Reforms (Washington: Center on Budget and Policy Priorities, 2000).
3. J.Henneberry, State Tools to Provide Family Health Insurance Coverage (Washington: National Governors Association, 4 January 1999).
4. G.Engquist and P. Burns, Health Insurance Flexibility and Accountability Initiative: Opportunities and Issues for States, State Coverage Initiatives Issue Brief Vol. 3, no. 2 (Washington: AcademyHealth, August 2002).
5. See, for example, C.Schoen and C. DesRoches, "Uninsured and Unstably Insured: The Importance of Continuous Insurance Coverage," Health Services Research 35, no. 1, Part II (2000): 187–206[Web of Science][Medline]; and L.Ku and D.C. Ross, Staying Covered: The Importance of Retaining Health Insurance for Low-Income Families, Report prepared by the Center on Budget and Policy Priorities (New York: Commonwealth Fund, 2002).
6. H.S.Farber, Mobility and Stability: The Dynamics of Job Change in Labor Markets, Princeton Industrial Relations Section Working Paper no. 400 (Princeton, N.J.: Princeton University, 1998).
7. Institute for Health Policy Solutions, Effective Coverage Expansions for Uninsured Kids and Their Working Parents: Links to Job-Based Coverage, Conference Transcript (Washington: IHPS, 18 May 2001).
8. J.Cohen, Design and Methods of the Medical Expenditure Panel Survey Household Component, MEPS Methodology Report 1 (Rockville, Md.: Agency for Healthcare Research and Quality, 1997).
9. P.Cooper and B. Schone, "More Offers, Fewer Takers for Employment-Based Health Insurance: 1987 and 1996," Health Affairs (Nov/Dec 1997): 142–149.
10. S.H.Long and M.S. Marquis, "Low-Wage Workers and Health Insurance Coverage: Can Policymakers Target Them through Their Employers?" Inquiry (Fall 2001): 331–337.
11. J.A.Klerman, J.L. Buchanan, and A. Leibowitz, Labor Turnover and Health Insurance in Health Benefits and the Workforce (Washington: U.S. Department of Labor, Pension and Welfare Benefits Administration, 1992).
12. We fit discrete-period models rather than continuous-duration models because a number of analysts have observed "seam" bias in reporting transitions; that is, most transitions that are measured occur between the end of one interview period and the beginning of the next. J.A. Klerman, "Pitfalls of Panel Data: The Case of the SIPP Health Insurance Data; The 1990s: A Decade of Decisions for Vital and Health Statistics," Proceedings of the 1991 Public Health Conference on Records and Statistics (Huntsville, Md.: U.S. Department of Health and Human Services, 1992), 36–39. We chose six-month intervals to reduce seam bias because the MEPS interview period was six months.
13. We restricted our analysis to those in the full panel so that our loss rates over time would not be influenced by a changing sample and to use the constant, full-panel weights for each case. We examined parents’ six-month job survival rates for children who survived to the second wave of each panel and found only small differences from the full panel. In MEPS 14.8 percent of jobs in the larger sample ended within six months, compared with 14.2 percent in the full panel; in SIPP the numbers were 12.0 percent and 10.5 percent, respectively.
14. The final six-month period is not included because many observations are censored—that is, we were unable to determine whether subjects had left their job before completing the panel.
15. Group coverage, new job, parent’s age, and industry were statistically significant. In addition, in the SIPP model, wage, occupation, union status, parent’s education, and child’s age were statistically significant (p = .05).
16. We terminated a Medicaid episode when there was a reported gap in coverage of at least one month. In the MEPS sample 8 percent of those with a new Medicaid episode had multiple episodes, and the average lapse in coverage was 4.9 months. About 30 percent had multiple episodes in SIPP because it is longer panel; the average lapse in coverage was 7.3 months.
17. In MEPS 93 percent were reported to be Medicaid episodes. Analyses using only Medicaid episodes provide similar results. In SIPP we cannot distinguish between Medicaid and other public insurance.
18. Essentially, we are reweighting observed Medicaid durations to represent the full sample eligible for premium assistance.
19. We also examined whether the job durations for the parent of a Medicaid child eligible for premium assistance differ from those of a child who does not enroll in Medicaid, because this difference might also bias our conclusions. We did not find differences.
20. L.Dubay, I. Hill, and G.M. Kenney, Five Things Everyone Should Know about SCHIP, New Federalism: Issues and Options for States, no. A-55 (Washington: Urban Institute, 2002).
21. We limited this analysis to SIPP because it is a longer panel. However, SIPP represents conditions in prosperous economic conditions, and job gaps in today’s economic climate could be different.
22. Our analysis includes only employees in small firms that offer insurance. Conclusions about the relationship between firm size and duration for all jobs could be different.
23. The time between interviews is longer in MEPS than in SIPP; some of the difference in duration could be attributed to differential seam bias and to failure to detect short episodes within a recall period. Differences in sample size and the measure of insurance offer also could contribute to the relative variability of the estimates.
24. Henneberry, State Tools to Provide Family Health Insurance Coverage. Decreased job mobility is not necessarily positive. Job mobility is necessary for enhanced productivity. See J. Gruber, "Health Insurance and the Labor Market," in Handbook of Health Economics, Vol. 1A, ed. A.J. Culyer and J.P. Newhouse (Amsterdam: Elsevier, 2000).
25. The MEPS data only provide annual income and so preclude such a model.
26. AcademyHealth, "State Health Coverage Matrix," September 2002, www.statecoverage.net/matrix.htm (11 June 2003).
27. AcademyHealth, Employer Buy-In Programs: How Four States Subsidize Employer-Sponsored Coverage (Washington: AcademyHealth, 2001); and AcademyHealth, State of the States Report (Washington: AcademyHealth, 2002).

                      What's this?

This article has been cited by other articles:

				M. D. Kogan, P. W. Newacheck, L. Honberg, and B. Strickland Association Between Underinsurance and Access to Care Among Children With Special Health Care Needs in the United States Pediatrics, November 1, 2005; 116(5): 1162 - 1169. [Abstract] [Full Text] [PDF]

				J. B. Mitchell, S. G. Haber, and S. Hoover Premium Subsidy Programs: Who Enrolls, And How Do They Fare? Health Aff., September 1, 2005; 24(5): 1344 - 1355. [Abstract] [Full Text] [PDF]