Health Affairs

This Article Abstract Submit a response to this article Services E-mail this article to a friend Alert me to new issues of the journal

H E A L T H  T R A C K I N G :
M A R K E T W A T C H 
 
1 November 2005  
  

Changes In Economic Conditions 
And Health Insurance Coverage, 
2000–2004

Coverage among adults worsened, but increases
in public insurance kept children’s coverage rates high.


by John Holahan and Allison Cook


ABSTRACT:

Between 2000 and 2004, the number of uninsured Americans increased by six million, primarily because of a decline in employer-sponsored insurance. All of the increase occurred among adults, for whom the drop in employer coverage was not offset by an increase in public coverage. The number of uninsured children fell slightly. About two-thirds of the growth in the uninsured was among Americans below 200 percent of the federal poverty level. Coverage rates have also fallen among higher-income Americans. About half of the growth in the uninsured was among young adults ages 19–34, about 55 percent among whites, and 73 percent among native-born citizens.

On 30 August 2005 the Census Bureau released data showing that the number of uninsured Americans had increased by 850,000, to 45.8 million, between 2003 and 2004. The Census Bureau also provided data that showed that the U.S. poverty rate increased from 12.5 percent to 12.7 percent, that real median household incomes were unchanged, and that real median earnings of both men and women who work full time had declined.¹ The number of uninsured adults actually increased by 950,000, while the number of children declined by 100,000. Both adults and children saw a decline in employer-sponsored insurance coverage. The decline in coverage for children was offset by strong growth in public coverage, but this was not the case for adults.

These changes continued trends that have been present since the beginning of the decade. In this paper we focus on the changes in coverage that occurred between 2000 and 2004—years in which the U.S. economy was in a recession followed by a period of very slow economic growth. During this period, the U.S. poverty rate increased from 11.3 percent to 12.7 percent and real median income fell from $46,058 to $44,389. Our findings show how the number of uninsured people has changed in response to the decline in employer coverage as well as to changes in employment and the distribution of income.

Study Data And Methods

We used data from the 2001 and 2005 March Supplements to the Current Population Survey (CPS). The CPS is the most frequently cited national survey on Americans’ health insurance. Its strengths and weaknesses have been widely documented.² There is a debate over whether the CPS is measuring the uninsured for an entire year (as intended) or whether responses more closely reflect the uninsured at a point in time. In this paper we assume that the CPS is essentially measuring point-in-time coverage, primarily because the number of uninsured people in the CPS has historically been closer to point-in-time estimates and well above other surveys’ full-year uninsurance estimates. In its most recent release, the Census Bureau commented on this issue and stated that its estimates were more closely in line with point-in-time estimates.³

There is also an issue of whether the CPS understates Medicaid enrollment and possibly overstates the number of uninsured people. Although a Medicaid undercount might lead to an overstatement of the number of uninsured people in the CPS, recent evidence seems to suggest that it might be primarily overstating private coverage.⁴ Nonetheless, the CPS provides the best available measure of changes in coverage over time because modifications to the survey have been infrequent.

The CPS allows respondents to report multiple types of coverage. In this analysis, responses are classified in a hierarchy, with each respondent assigned only one type of coverage: the type of coverage that is highest in the hierarchy. In descending order, the hierarchy is as follows: employer, Medicaid, State Children’s Health Insurance Program (SCHIP), or state coverage (it is not possible to distinguish SCHIP from Medicaid coverage in the CPS); military, veterans, or Medicare coverage; directly purchased coverage; and uninsured.

The annual CPS supplement on health insurance uses household income. In this paper we use the income of the health insurance unit (HIU), which includes members of the nuclear family who can be covered under one policy (policyholder, spouse, children under age nineteen, and full-time students under age twenty-three).⁵ Counting the income of all members of the household can overstate family income because it includes the income of all relatives and unrelated people living together.⁶ The income of the HIU more accurately reflects the income available when purchasing private insurance or determining eligibility for public programs.

In this paper we present data for three income groups using poverty thresholds: those with incomes below 200 percent of the federal poverty level (low), 200–399 percent (middle), and 400 percent and higher (high).⁷ An advantage of using poverty thresholds is that they adjust for family size and inflation. The Census Bureau’s household income categories adjust for neither one, and they have led to some misinterpretation of data in the Census reports.⁸

Changes In Coverage, 2000–2004

Changes in employment and incomes. Between 2000 and 2004, the Bureau of Labor Statistics (BLS) Establishment Survey showed a decline of 2.6 million jobs through 2003, although technically the nation was in recession only from March 2001 until November 2001.⁹ Between 2003 and 2004, there was an increase of 1.5 million jobs. But as we have noted above, median household incomes declined, and the poverty rate increased throughout this period. This period is instructive because it shows how much the U.S. health insurance system is affected by underlying economic conditions.

Although the overall number of people in working households rose during this period, fewer families had two full-time workers; there was a shift toward families with one full-time or only part-time workers, and more people lived in households with no worker (Exhibit 1). These shifts are important because rates of employer coverage decline and uninsurance rates rise as the number of full-time workers in the households declines. Also, while the population grew by ten million, there was an increase of 9.1 million people with incomes below 200 percent of poverty. These economic trends had important implications for changes in the patterns of health insurance coverage.

Changes in coverage. Nonelderly. Between 2000 and 2004, the number of uninsured nonelderly Americans increased by 6.0 million (Exhibit 2). Employer-sponsored coverage fell 4.6 percentage points; Medicaid and state-sponsored insurance increased 2.4 percentage points. The small increases in Medicare/TRICARE and directly purchased insurance only partially offset the decline in employer coverage, and the uninsurance rate increased from 16.1 percent to 17.8 percent.

About two-thirds of the increase in nonelderly uninsured people was among those below 200 percent of poverty. This income group saw a sharp drop in employer coverage (from 34.7 percent to 30.0 percent) that was not wholly offset by an increase in public coverage; their uninsurance rate increased from 31.5 percent to 32.6 percent. However, the most important factor was that the population below 200 percent of poverty increased by 9.1 million. Most of this increase was among those below 100 percent of poverty (data not shown). Thus, people moved into economic circumstances in which the likelihood of employer coverage was less and uninsurance rates were higher.

Also during this period, the number of uninsured middle- and higher-income Americans rose, by 1.3 million and 0.7 million, respectively. Both income groups saw declines in employer coverage that were not offset by increases in other forms of coverage. The increases in uninsurance rates in each income group, coupled with the general shift of the population toward more low-income people, contributed to the overall increase in the uninsurance rate of 1.7 percentage points and thus to the increase of six million uninsured people.

Adults. All of the increase in the uninsured was among adults (Exhibit 3). The adult population increased by 8.5 million between 2000 and 2004, while the number of uninsured adults increased by 6.3 million. Employer coverage among adults declined 4.5 percentage points. Because public coverage is less available to adults, the small increase in public coverage that occurred did not offset the decline in employer coverage. Thus, the uninsurance rate rose from 17.9 percent to 20.6 percent. Again, about two-thirds of this increase (4.2 out of 6.3 million) was among people below 200 percent of poverty. Because of employer coverage declines, the uninsured rate among low-income U.S. adults increased from 37.6 percent to 40.3 percent (Exhibit 2). Employer coverage also declined among middle- and high-income adults, which led to increases in their uninsurance rates.

Although the bulk of the increase in uninsurance was among low-income adults, there were increases for higher-income adults as well (Exhibit 2). Again, the rising uninsurance rates among all income groups, together with the shift of the population toward more low-income adults, led to the overall increase in the number of uninsured adults between 2000 and 2004.

Children. Children fared far better than adults (Exhibit 4). They had comparable declines in employer coverage—4.6 percentage points overall. But these were more than offset by increases in public coverage. Thus, the uninsurance rate among children actually declined, from 12.3 percent to 11.6 percent, and the number of uninsured children declined by 350,000. Among low-income children, the role of public coverage is particularly striking. Although the rate of employer coverage declined from 36.1 percent to 30.6 percent, there was an increase in public coverage from 36.5 percent to 44.7 percent. Uninsurance declined 2.3 percentage points, in contrast to adults, for whom the rate rose 2.7 percentage points. There were only small changes in the number of uninsured middle- and higher-income children.

One question that often arises is whether the drop in employer coverage for children was the result of the expansion of public programs (crowding out) or whether the availability of public coverage simply buffered the decline in employer coverage that occurred because of the economic slowdown. The sizable drop in employer coverage for both children and adults overall and for both low-income and middle-income adults and children suggests that the availability of Medicaid and SCHIP coverage was not likely to have been the predominant cause of the decline in employer coverage for children during this period. In other words, the decline in employer coverage for adults was comparable to that for children (slightly less for lower-income adults but higher for middle-income adults) despite the fact that the availability of public coverage was much less. This suggests that the drop in employer coverage, not the availability of public coverage, is the precipitating factor.

Age, income, and declining adult coverage. To further explore the decline in coverage among adults, we examined changes in income and insurance coverage among adults across various age groups (Exhibit 5). Among those ages 19–34, there was a large increase in the number below 200 percent of poverty. The number of young adults in both of the higher-income groups declined. In the 35–54 age group there also was a large increase in the number with incomes below 200 percent of poverty, with a slight increase in the middle-income group and a sharp decline in the highest-income group.

In contrast, for those ages 55–64, there was a small increase in the lowest-income group. There was a slightly larger increase in the number of middle-income and a sizable increase in the number of higher-income near-elderly people. Overall, because of the baby boom, the 55–64 age group increased by 4.9 million. Thus, the adult population shifted toward the age group more likely to have some form of insurance—either private or public—and less likely to be uninsured. In the absence of such a shift in age composition, the coverage picture would have been worse.

Almost half of the increase in the number of uninsured adults occurred in the group ages 19–34 (Exhibit 6). In this age group, 80 percent of the decline in coverage occurred among those below 200 percent of poverty (data not shown). Young adults experienced a sharp decline in employer coverage rates. Public coverage rates rose; so did the uninsurance rate (4.2 percentage points). As a result, three million more young adults lacked coverage.

Among Americans ages 35–54, there was also a decline in employer coverage and a sharp increase in the uninsurance rate (2.9 percentage points). About two-thirds of the increase in uninsurance in this age group occurred among those below 200 percent of poverty (data not shown), again largely because of the decline in employer coverage rates. Higher-income Americans in this age group also saw declines in employer coverage, as well as rising uninsurance. Overall, there was an increase of 2.7 million middle-aged Americans without insurance.

The largest growth in the adult population was among people ages 55–64. This group also experienced substantial gains in incomes despite the slow economy. The near-elderly saw no statistically significant change in employer coverage and consequently no significant change in the uninsurance rate. Because of population growth, the number of uninsured near-elderly Americans increased by 600,000. The increase in the number of uninsured people in this age group was equally split among the three income categories.

Changes in coverage among workers. To further explore the changes in adults’ coverage, we looked at how workers were affected. During this four-year period, the number of working Americans rose 1.4 million, according to the CPS (Exhibit 7). In the early part of this period, the number of workers declined, but this changed substantially during 2004. The BLS Establishment Survey shows an even larger decline in the number of workers between 2000 and 2003 and about the same increase between 2003 and 2004. This survey is generally regarded as more reliable for measuring employment trends because of its larger sample size. However, the CPS, despite having smaller samples of workers, captures self-employment and employment in new firms, which is important for understanding changes in insurance coverage.¹⁰

The number of workers without insurance increased by 3.8 million during the study period. Among those working, the rate of employer coverage declined 3.7 percentage points, which led to a 2.5-percentage-point increase in the uninsurance rate. Almost two-thirds of the increase occurred among workers.

Looking at the data by size of firm, one sees a shift from employment in midsize and large firms, which have relatively high employer coverage rates, to self-employment and employment in small firms, where employer coverage rates are lower. At the same time, rates of employer coverage in small firms and among the self-employed dropped sharply—more so than in midsize and large firms (Exhibit 7). This led to an overall increases in uninsurance rates among those working in small firms and among the self-employed. The result is that 2.6 million of the 3.8 million increase in the number of uninsured workers occurred among the self-employed and those in small firms.

Exhibit 7 provides the same data for industries with historically high and low rates of employer coverage.¹¹ Between 2000 and 2004, there was an employment decline of 2.1 million in high-coverage industries and an employment increase of 3.5 million in low-coverage industries. Thus, workers were shifting toward firms less likely to offer coverage. At the same time, the rate of employer coverage fell 4.2 percentage points in low-coverage industries and 1.6 percentage points in high-coverage industries. The decline in the employer coverage rate together with the shifts among workers led to an increase of 3.5 million uninsured workers in low-coverage industries.

Race/ethnicity, citizenship, and region. Finally, we examined changes in coverage by racial and ethnic group, citizenship, and region of the country (Exhibit 8).¹² More than half (55 percent) of the increase in the uninsured over this period occurred among whites. Whites had the largest increase in numbers falling below 200 percent of poverty. Although the white population increased by 0.7 million, the number of white Americans below 200 percent of poverty increased by 4.0 million. For other groups, the increase in the number below 200 percent of poverty was less than the overall growth in their numbers (data not shown). In part because of the decline in incomes, whites had the largest growth in uninsurance rates. Each of the other groups experienced declines in employer coverage and increases in public coverage. Because the increase in public coverage among whites was smaller than for other groups, whites had a larger uninsurance rate increase than other groups. Despite the increase, uninsurance rates among whites remained well below those of other groups: 13.2 percent (in 2004) versus 21.2 percent for blacks, 34.3 percent for Hispanics, and 18.8 percent for others.

Almost three-quarters of the growth in the uninsured was among native-born citizens (Exhibit 8). The growth in the number of uninsured noncitizens was disproportionate to their overall population growth, but nonetheless, noncitizens accounted for only 20 percent of the growth in the uninsured. Each group saw a decline in employer coverage that was offset by increases in Medicaid and state-sponsored coverage.

When we examined uninsurance by region of the country, we found that more than half of the growth in uninsurance occurred in the South. This region experienced by far the greatest growth in population, which contributed to the large uninsurance increase. But it also had a large reduction in employer coverage rates, with a decline from 65.2 percent to 60.1 percent, as well as the largest increase in uninsurance rates. Also, the South had a large increase in its low-income population. The other three regions saw sizable reductions in employer coverage rates, increases in public coverage, and significant increases in uninsurance rates (Exhibit 8).

Discussion And Conclusions

The four most important findings in this paper are as follows. (1) The number of uninsured Americans rose by 6.0 million between 2000 and 2004 (the U.S. population increased by 10.0 million). The increase in uninsurance occurred primarily because of the decline in employer coverage, which fell both because a large number of Americans were not working and because coverage declined among workers. The latter no doubt reflects increases in health insurance premiums, which rose much faster (12.2 percent per year) than wages (2.9 percent per year) during this period.¹³ The percentage of small and midsize businesses (3–199 workers) offering health benefits also declined, from 68 percent in 2000 to 63 percent in 2004).¹⁴ Employers seem to have tried to shift the cost of health insurance to workers, and it is likely that a growing share of workers chose not to accept employers’ offers.¹⁵

The change in coverage was affected also by the shift in employment from industries that historically have had high rates of coverage to industries that have not, as well as from large firms to small firms and self-employment. Employer coverage rates fell in all types of industries and firms, but the declines were particularly great in low-coverage industries and in small firms. The fact that employer coverage rates fell among workers, not just among those who lost their jobs, suggests that rates of coverage could continue to decline, even as the economy improves.

(2) About two-thirds of the increase in uninsurance was among people below 200 percent of poverty. This is due to increases in both the size of the low-income population and the uninsurance rate of that group. Importantly, however, middle- and higher-income Americans were also clearly affected. The remaining one-third of the growth in uninsurance (2.0 million) occurred among those above 200 percent of poverty, even though this group only grew by 900,000. Thus, the lack of health insurance coverage is clearly beginning to affect middle- and higher-income Americans.

(3) Much of the increase in the uninsurance occurred among young adults, whites, and the native-born. About 50 percent of the uninsurance growth was among those ages 19–34; about 55 percent among whites; and about 73 percent among native-born citizens. Thus, rising uninsurance is clearly not a problem affecting primarily racial and ethnic minorities and noncitizens. Further, more than half of the increase in the uninsured occurred in the South, where uninsurance rates were already the highest in the country.

(4) Children actually gained health insurance coverage. The expansions of coverage in Medicaid and SCHIP that occurred in the late 1990s meant that children’s coverage was maintained, even with the loss of employer coverage. The expansion of public programs increased enrollment substantially. The result was actually a slight decline in uninsurance among children. Adults’ experiences were in sharp contrast, primarily because adults experienced the same declines in employer coverage but did not have the same access to public coverage. Also, people ages 55–64 actually saw improvements in both income and health insurance coverage. Without the gains seen for this group, the overall picture of rising uninsurance would have been much worse.

The decline in employer coverage is likely to continue. Increases in health care costs, and thus health insurance premiums, are likely to continue to grow faster than workers’ earnings. The decline in employer coverage will be further exacerbated if the shift from working in large and midsize firms to small firms and self-employment and from high- to low-coverage industries continues. The problem of the uninsured can be addressed in many different ways, such as tax credits or public program expansions, but doing so is likely to prove very difficult. Federal budget deficits are large, which will limit the federal government’s ability to act for the foreseeable future, and all indications are that the government will pursue spending cuts to address the huge cost of hurricane recovery in the coming years.¹⁶ States also face serious budget problems in part because increases in health care spending outpace the growth in state revenues, a trend that is likely to continue. As a result, it is difficult to envision a reversal of the trends we have described in this paper.

The authors greatly appreciate the comments and suggestions provided by Catherine Hoffman of the Kaiser Commission on Medicaid and the Uninsured and Genevieve Kenney and Linda Blumberg of the Urban Institute. They gratefully acknowledge the Kaiser Commission for its continued support of this work.

NOTES

1. U.S. Census Bureau, Income, Poverty, and Health Insurance Coverage in the United States: 2004, August 2005, www.census.gov/prod/2005pubs/p60-229.pdf (5 October 2005).
2. K. Lewis, M. Ellwood, and J.L. Czajka, “Counting the Uninsured: A Review of the Literature,” Assessing the New Federalism Occasional Paper no. 8, July 1998, www.urban.org/UploadedPDF/ occ8.pdf (5 October 2005); and Congressional Budget Office, “How Many People Lack Health Insurance and for How Long?” May 2003,
www.cbo.gov/showdoc.cfm?index=4210&sequence=0 (5 October 2005).
3. Census Bureau, Income, Poverty, and Health Insurance Coverage.
4. K.T. Call et al., “Uncovering the Missing Medicaid Cases and Assessing Their Bias for Estimates of the Uninsured,” Inquiry 38, no. 4 (2001/2002): 396–408; and K.T. Call, “Cumulative Evidence: The Impact of Response Error on Survey Estimates of Uninsurance” (Paper presented at the State Health Access Data Assistance Center Conference, Washington, D.C., 5 May 2005).
5. The CPS permits calculation of HIU-level income data because these data contain person-level income data that can be summed to any larger unit (HIU, subfamily, family, and so forth).
6. J. Holahan, C. Hoffman, and M. Wang, “The New Middle-Class of Uninsured Americans—Is It Real?” Issue Paper, March 2003, www.kff.org/uninsured/loader.cfm?url=/commonspot/security/getfile.cfm&PageID=14321 (20 October 2005).
7. The federal poverty level was $17,603 for a family of four in 2000 and $19,307 in 2004.
8. Holahan et al., “The New Middle-Class.”
9. National Bureau of Economic Research, Report of the Business Cycle Dating Committee, 17 July 2003, www.nber.org/cycles/july2003.pdf (5 October 2005).
10. Kathleen Utgoff, Bureau of Labor Statistics, Statement before the Joint Economic Committee, U.S. Congress, 5 September 2003.
11. All industries are included. The cutoff between high and low was an employer coverage rate of 80 percent in 2000. High-coverage industries include mining; manufacturing; utilities; finance/ insurance/real estate; education; and public administration. Low-coverage industries include agriculture; construction; transportation; wholesale/retail; trade; information/communication; professional health and social services; art/entertainment; other services; and former military.
12. A problem in analyzing data on race over this period was that there were changes in the wording of the race question on the CPS. Through the March 2002 CPS, respondents were instructed to choose one race—white, black, and so on. In the March 2003–2004 supplements, respondents were allowed to respond by indicating multiple races. Relatively few chose to do so; thus, we expect any effect on the results to be small. We characterized respondents as “white” if they reported being white alone and as “black” if they reported being black alone. By counting only those reporting being white alone or black alone in 2003, we understate the growth in the white or black population (and with any type of—or lack of—coverage) because those who reported a combination in 2003 would be counted as “other.” Respondents did not have the option of doing so in 2000. Hispanics were identified in response to separate questions; in this analysis, they are reported as “Hispanic” regardless of race.
13. J. Gabel et al., “Health Benefits in 2004: Four Years of Double-Digit Premium Increases Take Their Toll on Coverage,” Health Affairs 23, no. 5 (2004): 200–209.
14. Henry J. Kaiser Family Foundation/Health Research and Educational Trust, Employer-Sponsored Health Benefits, 2000–2004 surveys.
15. L. Blumberg and J. Holahan, “Work, Offers, and Take-Up: Decomposing the Source of Recent Declines in Employer-Sponsored Insurance,” Health Policy Online no. 9, 17 May 2004,
www.urban.org/UploadedPDF/1000645_healthpolicyonline_no9.pdf (5 October 2005).
16. M.A. Fletcher and J. Weisman, “Bush Says Spending Cuts Will Be Needed,” Washington Post, 17 September 2005.

John Holahan (jholahan{at}ui.urban.org) is director of the Health Policy Center at the Urban Institute in Washington, D.C., Allison Cook is a research assistant there.

DOI: 10.1377/hlthaff.w5.498
©2005 Project HOPE–The People-to-People Health Foundation, Inc.