This Article Extract Figures Only Reprint (PDF) Submit a response to this article Alert me when this article is cited Alert me when eLetters 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 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 Google Scholar Google Scholar Articles by Glied, S. Articles by Stabile, M. Search for Related Content PubMed PubMed Citation Articles by Glied, S. Articles by Stabile, M. Related Collections Insurance - Employer-Based System

Health Tracking

TRENDS

Generation Vexed: Age-Cohort Differences In Employer-Sponsored Health Insurance Coverage

How should policymakers respond to this? That depends critically on whether these low rates are transitory or permanent. If young adults gain insurance coverage as they grow older, then policymakers need to develop models that serve them during their transition to employer-sponsored insurance. If, conversely, low coverage rates in young adulthood presage continuing low rates, policymakers need to develop not age-specific but cohort-specific models that address permanent problems.

Employer-sponsored coverage is an outcome of the functioning of both the labor market and the health insurance market. Prior studies have generally concluded that labor market changes, such as shifts in the distribution of employment across industries, have had only minor effects on coverage.¹ Most of the change in employer-sponsored insurance appears to be a consequence of increases in the overall cost of coverage and in the employee contribution requirement.²

Our focus here is on the extent to which low (or high) rates of coverage among young adults persist within cohorts over time. This approach parallels a large literature in labor economics that has examined the earnings experience of cohorts over time. Earnings in the labor market tend to grow with experience and to improve as the economy becomes more productive. Many researchers have found, however, that the unusually large U.S. population cohorts born during the baby boom experienced persistently depressed wages relative to cohorts that preceded them—even as they grew older and the economy prospered.³ Declines in earnings were sharpest for baby boomers who did not graduate from college. Despite economy wide improvements in productivity, these workers never did as well as prior generations.

In this paper we replicate the methods used in the analyses of cohort earnings to analyze changes in employer-sponsored health insurance rates along three lines: the effect of secular time trends, the effect of growing older and gaining labor-market experience, and the effect of birth cohort.

Data And Methods

 
We use data on persons ages twenty-five to fifty from the March demographic supplement of the Current Population Survey (CPS) for the years 1980–1999. The CPS is a national survey of approximately 50,000 households each month.⁴ In each year the CPS asks whether or not a respondent is covered by employer-sponsored health insurance and, if yes, whether this coverage was made available through a current or former employer (or that of a family member). We define having employer-sponsored coverage to include anyone covered by such coverage, whether in one’s own name or through a family member.

The CPS health insurance questions changed several times during the 1980–1999 period. In general, these changes had little effect on the employer-sponsored coverage questions.⁵ all of our analyses include a variable for each 1980–1999 CPS year. These dummy variables capture any residual effects of questionnaire changes that affect the entire population. All analyses are weighted using the CPS weights.

To compare changes in health insurance to changes in earnings, we repeat our analyses for health insurance using information on weekly wages. We adjust all wages for inflation to 1980 dollars using the all-items Consumer Price Index (CPI). ⁶

We construct cohorts in the CPS data by following successive age groups over time. For example, the cohort born in 1952 was thirty years old in 1982. In 1987 the members of that cohort were five years older. The experience of the group is then captured by the experience of thirty-five-year-olds in 1987. In general, the sample of thirty-five-year-olds in 1987 is representative of the same population as the sample of thirty-year-olds in 1982. We restrict the sample for regression to cohorts born between 1940 and 1965 so that we have at least ten observations for each cohort in the data.

In further analyses, we compare the experience of different education groups. We define these groups according to highest level of education completed: (1) did not complete high school; (2) received a high school diploma diploma but had no further education; (3) attended college or postsecondary technical training but did not receive a degree; (4) received a college degree or higher.

We perform the decomposition by computing linear regressions that include age components, time-trend components, and cohort components. We specify the age component by including terms for ages 26–30, ages 31–35, ages 36–40, ages 41–45, and ages 46–50. The age components are measured relative to twenty-five-year-olds.⁷ We are interested in learning whether the transition to employer coverage has changed over time. We test this by including a second set of terms for the same age groups in the period after 1988 (the midpoint of our sample). These terms show whether the effect of being ages 26–30 is different after 1988 than before 1989.⁸ We specify time-trend components by including a dummy variable for each CPS year included in our analysis. We measure the trend components relative to 1989. We specify the cohort components using a series of dummy variables for each birth cohort year. We measure cohort components relative to the cohort born in 1955.

The coefficients we estimate in these analyses describe relative coverage rates. Rather than listing them, we plot the coefficients in a series of graphs that describe trends over age, time, or cohort (after controlling for each of the other variables). The vertical axes of these graphs show the estimated difference in insurance coverage associated with different values of that component (age, time trend, or cohort) relative to the base category for that component.

Study Results

Top Data And Methods Study Results Explaining The Pattern:... Interpreting The Results Policy Implications NOTES

 
Changes in age profile. Exhibit 1 shows how levels of employer-sponsored health coverage change as people age. The exhibit shows two profiles, one for the period before 1989 and one for the period 1989 and beyond. The upward slope of both age profiles shows that levels of employer-sponsored health insurance increase as people age. However, examining the difference between the two profiles reveals that the rate at which health insurance coverage increases with age has changed over time. The post-1988 profile is steeper than the pre-1989 profile, which suggests that after 1988 employer-sponsored health insurance coverage levels increased more rapidly between the ages of twenty-five and thirty-four.⁹

View larger version (13K): [in this window] [in a new window] EXHIBIT 1 Age Profiles For Employer-Sponsored Health Coverage Before And After 1989

Changes over time. Exhibit 2 shows the pattern of employer-sponsored health coverage over time relative to 1988 levels. The pattern over time reflects changes in economic conditions and changes to the CPS health insurance questionnaire. The apparent jump in coverage between 1994 and 1995 almost certainly reflects concurrent questionnaire changes.¹⁰ The data also suggest that questionnaire changes between 1987 and 1988 had little effect in this population. These results results suggest that the decline in health insurance coverage over this period ended in 1994, after controlling for cohort and age effects.

View larger version (12K): [in this window] [in a new window] EXHIBIT 2 Effect Of Current Year On Employer-Sponsored Health Coverage, 1981–1999

View larger version (13K): [in this window] [in a new window] EXHIBIT 3 Effect Of Birth Cohort On Employer-Sponsored Health Coverage, 1940–1965

The time-series profile of coverage by education group, controlling for age and cohort effects, is described in Exhibit 4. Over time, the propensity of the most highly educated workers to obtain employer-sponsored health insurance coverage has remained flat.¹¹ By contrast, the least-educated workers have experienced a continuous decline in the propensity to hold employer-sponsored coverage over time. Comparing this graph to the time trend of the unemployment rate reveals that declines in the unemployment rate are correlated correlated with increases in coverage for all but the most-educated groups.¹² For the least educated, however, the effect of cyclical improvements has been swamped by the downward trend in coverage over time.

View larger version (21K): [in this window] [in a new window] EXHIBIT 4 Effect Of Current Year On Employer-Sponsored Health Coverage, By Education Group, 1982–1999

View larger version (26K): [in this window] [in a new window] EXHIBIT 5 Employer-Sponsored Health Coverage, By Cohort Of Birth And By Educational Attainment

Top Data And Methods Study Results Explaining The Pattern:... Interpreting The Results Policy Implications NOTES

We repeat our analysis to examine the age, time-trend, and cohort patterns on weekly wages. First, we find that the experience profile in earnings fully replicates the profile for health insurance. Second, we find a downward trend over time in earnings for less-educated workers. However, the decline in coverage rates is much steeper than the decline in wages. For college-educated workers, earnings rise over time, but coverage is flat. The difference in the time-trend patterns for earnings and health insurance across all groups reflects the rising cost of health insurance relative to earnings over time.¹³ We also find that health insurance is more sensitive to business-cycle fluctuations than are wages. This is not surprising, since health insurance is contingent on employment, and employment changes tend to be stronger than wage changes over a business cycle.

Third, we find, as does the prior literature, a decline in the earnings of cohorts of high school–educated workers over time.¹⁴ The decline in earnings for this group is less pronounced than is the decline we find in the health insurance data. One explanation for the more pronounced decline in benefits than in wages for high school–educated cohorts born after 1950 may be the decline in unionization. ¹⁵ Union jobs are much more likely to offer benefits than are nonunion jobs, and high school–educated workers are more likely than other groups are to have been affected by the decline in private-sector unionization.

As in the health insurance data, there is little pronounced cohort pattern for workers with less than a high school education. In the wage data, in contrast to the health insurance data, cohorts of college-educated workers born after the mid-1950s show consistent increases in wages. This divergence suggests that persons in this group are most likely to be able to take advantage of market reforms that make it easier to buy health insurance.

Interpreting The Results

Top Data And Methods Study Results Explaining The Pattern:... Interpreting The Results Policy Implications NOTES

 
The results described above can best be understood by considering their implications for the health insurance experiences of different groups. Consider, for example, the group of high school–educated adults born in 1956. The age pattern for this group suggests that their health coverage should have increased by four percentage points between ages twenty-five and thirty-five. The time-trend component, however, shows that employer-sponsored insurance for high school– educated adults fell six percentage points between 1981 (when the cohort born in 1957 were twenty-five years old) and 1991 (when they were thirty-five). The combination of these two effects meant that for this cohort, insurance coverage rates at age thirty-five were actually lower than their rates were at age twenty-five. Rather than gaining employer- sponsored insurance coverage with age, this cohort actually lost it.

Now consider the high school–educated cohort born seven years later, in 1963. When this cohort turned twenty-five (in 1988), economic conditions overall for high school– educated workers were much as they were five years earlier. Nonetheless, this later cohort experienced a substantial decline in economic well-being and health insurance coverage compared with the 1956 cohort. At age twenty-five, their health insurance rates were fully 4.2 percentage points lower than those of their older counterparts were. The time series profile for high school–educated workers declined over the next ten years, so that, again, this cohort had lower rates of coverage at age thirty-five than at age twenty-five.

Policy Implications

Top Data And Methods Study Results Explaining The Pattern:... Interpreting The Results Policy Implications NOTES

 
The existence of age-related and cyclical components of coverage has several implications for policy. First, young adults move into coverage as they grow older. Policies that make it easier for young adults to make the transition to employer-sponsored coverage could flatten the experience profile. Second, coverage rates move with the business cycle. These fluctuations are sharpest for poorly educated workers. Policies that offer ways for these workers to obtain or maintain coverage during downturns could smooth cyclical patterns.

The results described above also document, however, that there is a strong permanent component to changes in employer-sponsored health coverage over the past two decades. This affects health insurance planning decisions. For example, projections of the number of older Americans who will be uninsured are likely to be more accurate if they build on cohort patterns than if they extrapolate from the experience of today’s sixty year-olds.¹⁶

The permanent component is strongest for high school–educated workers, who also experienced a sharp decline in earnings over this period. For more recent cohorts, those born since 1963, this decline appears to have ended; rates have improved slightly but remain well below those of a generation earlier. The existence of this permanent component implies that even if economic conditions and health care costs remain relatively stable, future rates of employer-sponsored insurance coverage for these cohorts are likely to remain well below those of previous cohorts.

This implies that more expansive policies will be needed to sustain increases in health coverage. Rather than targeting an age group or a group undergoing a distinct economic transition, policies need to make insurance available for extended periods to people defined by income, not age or employment status. This coverage might be offered through employers or through other venues that we have not considered in this paper, such as private nongroup or public coverage.¹⁷ The depressed earnings of this group suggest that the private nongroup market is unlikely to be a viable source of coverage, unless substantial subsidies are offered. Public coverage might be an option, but this coverage is generally not available to working adults ages twenty-five to fifty. Subsidies that enable workers to purchase coverage through their own employers also may help members of cohorts with poor health insurance prospects.

Finally, the decline in employer-sponsored health insurance coverage across cohorts of workers with only a high school education should be considered when the overall economic well-being of this group is being evaluated. For these Americans, declines in insurance coverage exacerbate an already negative trend in income.

Editor's Notes

 
Sherry Glied is associate professor and head of the Division of Health Policy and Management at Columbia University’s Mailman School of Public Health. Mark Stabile is assistant professor in the Department of Economics at the University of Toronto.

This research was supported by an Investigator Award (Glied) from the Robert Wood Johnson Foundation. The authors thank Dan O’Flaherty for his helpful comments.

NOTES

Top Data And Methods Study Results Explaining The Pattern:... Interpreting The Results Policy Implications NOTES

 

1. G. Acs, "Explaining Trends in Health Insurance Coverage between 1988 and 1991," Inquiry 32, no. 1 (1995): 102–110; and D. Chollet, "Employer-Based Health Insurance in a Changing Work Force," Health Affairs (Spring I 1994): 315–326.
2. R. Kronick and T. Gilmer, "Explaining the Decline in Health Insurance Coverage: 1979–1995," Health Affairs (Mar/Apr 1999): 30–47; S. Glied and M. Stabile, "Explaining the Decline in Insurance Coverage among Young Men," Inquiry 37, no. 3 (2000): 295–303; S. Long and M.S. Marquis, "The Uninsured Access Gap: Narrowing the Estimates," Inquiry 31, no. 4 (1994): 405–414; and P.F. Cooper and B.S. Schone, "More Offers, Fewer Takers for Employment-Based Health Insurance: 1987 and 1996," Health Affairs (Nov/Dec 1997): 142–149.
3. F. Welch, "Effects of Cohort Size on Earnings: The Baby Boom Babies’ Financial Bust," Journal of Political Economy 87, no. 5 (1979): S65–S98; F. Levy, The New Dollars and Dreams (New York: Russell Sage, 1998); and D. Card and T. Lemieux, "Can Falling Supply Explain the Rising Return to College for Younger Men? A Cohort-Based Analysis," NBER Working Paper no. 7655 (Cambridge, Mass.: National Bureau of Economic Research, April 2000).
4. We use uniform CPS files produced by Unicon and include only the outgoing rotation groups (so as to include each individual only once).
5. Glied and Stabile, "Explaining the Decline."
6. We compute weekly wages in the CPS by dividing annual wage and salary income by weeks worked in the prior year. We omit all observations with estimated wages below $50 perweek.
7. Using an alternative base age does not affect the relative shape of the relationship.
8. We selected 1988 simply because it is the midpoint and would yield the maximum number of observations to identify the trend. Our specification is nearly fully saturated, and we need many observations to identify each coefficient.
9. The F-test for inclusion of the interaction terms between the post-1988 dummy variable and the age effects is significant at p < .008.
10. In 1995 the order of CPS questions changed in a way that may have increased the count of persons with employer-sponsored health insurance.
11. The effects of the current year on changes in employer-sponsored health coverage are not statistically significant with the exception of 1993.
12. Council of Economic Advisers, Economic Report of the President (Washington: U.S. Government Printing Office, 1999).
13. Kronick and Gilmer, "Explaining the Decline."
14. Levy, The New Dollars and Dreams.
15. K. Blackburn et al., "The Declining Economic Position of Less-Skilled American Men," in A Future of Lousy Jobs? The Changing Structure of U.S. Wages, ed. G. Burtless (Washington: Brookings Institution, 1990), 31–67.
16. S. Glied and M. Stabile, "Covering Older Americans: Forecast for the Next Decade," Health Affairs (Jan/Feb 1999): 208–213.
17. For a discussion of changes in public coverage over this period, see L. Ku and B. Garrett, "How Welfare Reform and Economic Factors Affected Medicaid Participation: 1984–96," Assessing the New Federalism Discussion Paper 00–01 (Washington: Urban Institute, 2000).

                      What's this?