Over the pa st tw o de cad es the

Transcription

1 Generation Vexed: Age-Cohort Differences In Employer-Sponsored Health Insurance Coverage Even when today s young adults get older, they are likely to have lower rates of employer-related health coverage than their elders have today. b y S h e r ry G l i e d a n d M a rk S t a b i l e 184 Over the pa st tw o de cad es the rate of employer-sponsored health insurance coverage has dropped markedly, although the decline has not been smooth. This downward trend has been concentrated among young adults. Today, unlike twenty years ago, there is a substantial gap in employer-sponsored coverage between young adults and their elders. 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 labormarket 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 economywide 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. 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. H E A L T H A F F A I R S ~ V o l u m e 2, N u m b e r 1 21 Project HOPE ThePeople-to-PeopleHealth Foundation, Inc.

2 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 The CPS is a national survey of approximately 5, 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 period. In general, these changes had little effect on the employer-sponsored coverage questions. 5 All of our analyses include a variable for each 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 198 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 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 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 We restrict the sample for regression to cohorts born between 194 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 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 3, ages 31 35, ages 36 4, ages 41 45, and ages 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 3 is different after 1988 than before 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 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 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 n 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 185 H E A L T H A F F A I R S ~ J a n u a r y / F e b r u a r y 2 1

3 EX HIBIT 1 Age Profiles For Employer-Sponsored Health Coverage Before And After Before and beyond Age (years) SOURCE: Regression coefficients from analysis of the Current Population Survey that includes year and cohort effects. a Percent change in employer-sponsored coverage relative to twenty-five-year-old before 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 The shift in the age profile suggests that some of the decline in coverage experienced by today s younger adults will be made up as they grow older. Holding other factors constant, group coverage of today s twenty-fiveyear-olds is likely to increase nearly eight percentage points by the time they reach their mid-thirties. The changing age profile, however, does not fully explain the changes in health insurance coverage over time. n 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. 1 The data also suggest that questionnaire changes between 1987 and 1988 had little effect in this population. These results suggest that the decline in health insurance coverage over this period ended in 1994, after controlling for cohort and age effects. n Effects of cohort of birth. Finally, we examine the effects of cohort of birth on the pattern of employer-sponsored health coverage. Exhibit 3 shows the percentage-point change in coverage relative to the cohort born in Health insurance coverage rates peaked for the cohort born in 1947, just as the baby boom began. There has been a steady decline in coverage for subsequent cohorts. At each age, health insurance coverage rates for the cohort born in 1947 have been about three percentage points higher than rates for the cohort born in n Breakdown by education groups. Next, we repeat the analyses above for each of the four education groups listed earlier. We find a large and statistically significant change in the age profile of employer-sponsored health coverage after 1988 for the leasteducated group (controlling, once again, for year and cohort effects), a small but significant change for those with college education, and a small and insignificant change in the age profile for groups with high school and some H E A L T H A F F A I R S ~ V o l u m e 2, N u m b e r 1

4 EX HIBIT 2 Effect Of Current Year On Employer-Sponsored Health Coverage, CPS year SOURCE: Regression coefficients from analysis of the Current Population Survey that includes age and cohort effects. a Percent change in employer-sponsored coverage relative to college education. In the less-educated groups the age profile becomes slightly steeper. The difference between coverage rates at age twenty-five and at age thirty-five is about six percentage points greater in the later period for those with less than a high school education and about two percentage points greater for those with a high school education. For groups with a high school education or less, however, the entire age profile, while becoming steeper, also shifts downward. This shift is so large that the earlier and later profiles do not cross within the sample age range. 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 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. Exhibit 5 illustrates cohort patterns of coverage for these four education groups, controlling for age and year effects. Health insurance coverage has remained relatively flat in cohorts born since 195 among those in each group except those with a high school diploma. Cohorts of high school educated workers born since 1947 have seen steady declines in their propensity to hold health insurance, with a slight uptick in the youngest cohorts in the sample. Explaining The Pattern: Comparison To Wage Effects In designing policy, it would be useful to know whether health insurance coverage patterns parallel changes in earnings. If insurance coverage and earnings have moved separately, policy changes in the health insurance market might enable cohorts with poor employer- 187 H E A L T H A F F A I R S ~ J a n u a r y / F e b r u a r y 2 1

5 EX HIBIT 3 Effect Of Birth Cohort On Employer-Sponsored Health Coverage, Cohort of birth SOURCE: Regression coefficients from analysis of the Current Population Survey that includes year and cohort effects. a Percent change in employer-sponsored coverage relative to 1955 birth cohort sponsored coverage to purchase their own insurance. If, however, earnings declines have paralleled coverage declines, cohorts without coverage are also likely to lack the earnings to take advantage of market reforms. 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 195 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 H E A L T H A F F A I R S ~ V o l u m e 2, N u m b e r 1

6 EX HIBIT 4 Effect Of Current Year On Employer-Sponsored Health Coverage, By Education Group, Post high school but no degree College degree or more Less than high school diploma High school diploma CPS year SOURCE: Regression coefficients from separate analyses of the Current Population Survey by education category that each include age and cohort effects. NOTE: Coverage rates among persons age thirty-five in each year. a Percent change in employer-sponsored coverage relative to wage data, in contrast to the health insurance data, cohorts of college-educated workers born after the mid-195s 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 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 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 employersponsored 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 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 timeseries 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 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 189 H E A L T H A F F A I R S ~ J a n u a r y / F e b r u a r y 2 1

7 EX HIBIT 5 Employer-Sponsored Health Coverage, By Cohort Of Birth And By Educational Attainment Post high school but no degree Less than high school diploma 1-1 College degree or more High school diploma Cohort of birth SOURCE: Regression coefficients from separate analyses of the Current Population Survey by education category that each includes age and year effects. a Percent change in employer-sponsored coverage relative to cohort born in 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 employersponsored 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 sixtyyear-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 H E A L T H A F F A I R S ~ V o l u m e 2, N u m b e r 1

8 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. 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 1. G. Acs, Explaining Trends in Health Insurance Coverage between 1988 and 1991, Inquiry 32, no. 1 (1995): 12 11; and D. Chollet, Employer- Based Health Insurance in a Changing Work Force, Health Affairs (Spring I 1994): R. Kronick and T. Gilmer, Explaining the Decline in Health Insurance Coverage: , Health Affairs (Mar/Apr 1999): 3 47; S. Glied and M. Stabile, Explaining the Decline in Insurance Coverage among Young Men, Inquiry 37, no. 3 (2): ; S. Long and M.S. Marquis, The Uninsured Access Gap: Narrowing the Estimates, Inquiry 31, no. 4 (1994): ; 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): 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 (Cambridge, Mass.: National Bureau of Economic Research, April 2). 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 $5 per week. 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 < 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 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 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, 199), S. Glied and M. Stabile, Covering Older Americans: Forecast for the Next Decade, Health Affairs ( Jan/Feb 1999): 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: , Assessing the New Federalism Discussion Paper -1 (Washington: Urban Institute, 2). 191 H E A L T H A F F A I R S ~ J a n u a r y / F e b r u a r y 2 1