Current Issue Pic Top Harvard Health Policy Review Archives Top
Current Issue Pic Middle About Us Fill Archives Bottom
Contact Us Fall 2000; Volume 1, Number 1
In Focus

Understanding Trends in Employment-based Health Insurance Coverage
Patricia Keenan
page 1 | page 2

Many have documented that employment-based coverage has declined since the 1980s. However, coverage levels have been increasing in recent years. As the primary source of health insurance coverage for non-elderly Americans, employment-based coverage has a large impact on trends in rates of coverage and lack of insurance in the United States. This paper compiles the results from analyses of several surveys to assess trends in employment-based coverage since the 1980s and possible explanations for coverage trends. Trends in coverage are reviewed first, differentiating by time period and among population subgroups. Next, potential explanations for the decline in insurance coverage between the 1980s and 1990s are identified and evaluated. The paper concludes with a brief discussion of the implications of research for health policy. It shows that while coverage declined between the 1980s and 1990s, particularly among low-income workers, overall, employment-based coverage has increased since 1994. Furthermore, rising health insurance premiums and slow wage growth, rather than labor market shifts or crowd out, are cited as the primary reasons for the decline in coverage between the 1980s and 1990s. Comparable evidence is not available to assess the factors behind the recent upturn in employment-based coverage.

Trends in Employment-based Health Insurance Coverage
By all accounts, the share of the non-elderly population covered by employment-based health insurance has declined since the 1980s.1 Between 1987 and 1996, coverage rates for the non-elderly population declined by 6 percentage points, from 70.9 percent to 64.1 percent (Gabel, 1999). Another set of estimates suggest that the decline since 1979 is greater. From 1979 to 1997, coverage for adult workers (ages 20 to 64) declined by 7.4 percentage points (Farber and Levy, 2000).

Some groups have experienced relatively larger declines in employment-based coverage. Between 1989 and 1993, employment-based coverage among children declined from 62.7 percent to 56.5 percent, or 6.2 percentage points, reflecting a decline in dependent coverage (Holahan and Kim, 2000). Among adults (both employees and dependents), coverage rates declined by 5.0 percentage points, to 62.4 percent in 1993 (Holahan and Kim, 2000). Furthermore, trends by wage show a greater effect on low wage (< $7.00 per hour) as compared to higher wage (>$7.00 per hour) workers. Low wage workers experienced a 6.9 percentage point decline in employment-based coverage between 1987 and 1996, while rates of coverage for higher wage workers remained essentially flat (Monheit and Schone 2000). In addition, the gap in coverage for adult workers with and without a college degree widened between 1979 and 1997 (Farber and Levy, 2000).

Assessing trends between the 1980s and 1990s, however, masks increases in coverage that have occurred within the 1990s. Estimates of employment-based coverage for adults and workers from the March CPS (Current Population Survey), the CPS Supplements, and the SIPP (Survey of Income and Program Participation) show increases in coverage since the mid 1990s (Currie and Yelowitz, 1999; Farber and Levy, 2000; Long and Marquis 1999; Fronstin 2000, Holahan and Kim 2000). Because income is a strong predictor of employment-based health insurance coverage, as incomes have risen, rates of employment-based coverage have also increased in this period (Holahan and Kim, 2000). Among the non-elderly, rates increased by 1.5 percentage points, to 65.8 percent, between 1994 and 1998 (Holahan and Kim, 2000). This trend extends to vulnerable subgroups. Employment-based coverage among children increased 2.7 percentage points between 1994 and 1998 (Holahan and Kim, 2000). Furthermore, rates of employment-based coverage increased slightly, by 0.7 percentage points among those with income below 200 percent of the poverty level (although this increase reflects a reduction in the number of people in this income category, not an increase in the number insured). Coverage rates decreased among the 200 to 399 percent of poverty and the 400+ percent of poverty groups, by 1.5 percentage points and 0.9 percentage points, respectively. Within the 400+ percent group, a gain in health insurance coverage by 11 million individuals between 1994 and 1998 did not outweigh an overall growth in this group of 13 million persons in this time period. Assessing these trends by worker education level reveals that although the differential in coverage between workers with and without a college education increased from 6.3 percentage points to 13.4 percentage points between 1979 and 1993, it narrowed to 12.9 percentage points by 1997 (Farber and Levy, 2000).

To better understand the mechanics behind coverage trends, rates of coverage can be broken into three steps: whether the employer offers coverage, whether the employee is eligible for coverage, and finally, whether eligible employees take up coverage. Analyses of these components show that declines in employment-based coverage are largely attributable to declines in employee takeup of health insurance, though less skilled workers have faced reductions in offer rates and eligibility as well. Between 1987 and 1996, employer offers of health insurance increased by three percentage points, while employee takeup declined by eight percentage points (Cooper and Schone, 1997). Farber and Levy (2000) find that declines in takeup by workers in long term full-time jobs and declines in eligibility for part-time workers are the major contributors to a 4.4 percentage point decline in employment-based insurance coverage between 1988 and 1997. Again, notably, the results show that employer offers of coverage actually increased over this period.2

Assessments of these trends by worker income indicate departures from this overall trend. Estimates of whether workers have "access" to coverage, i.e., whether a worker is eligible for coverage through their own or a family member's job, show increases in the access rates for high wage workers, but a decline among low wage workers. Between 1987 and 1996, the "access" rate increased from 92.2 percent to 96.1 percent for high wage (>$15.00/hr) workers, but declined by 4.9 percentage points for low wage workers, from 60.3 to 55.4 percent (Cooper and Schone, 1997, see also O'Brien and Feder, 1998). Furthermore, while takeup rates for workers with access declined only slightly for high wage workers (from 94.1 percent in 1987 to 93.9 percent in 1996), takeup rates declined by 13.6 percentage points for low wage workers over this time period.

Similarly, trends among the less educated reflect reductions in employer offer and eligibility, as well as takeup. Among workers with less than a high school education, the 8.6 percentage point decline in coverage between 1988 and 1997 is attributable to reductions in offer, eligibility, and takeup (Farber and Levy, 2000). For college graduates, the much lower 2.9 percentage point decline in coverage is primarily attributable to reductions in takeup and, to a lesser extent, in eligibility. Among college graduates, availability of spousal coverage offsets some of the decline in coverage, while for those with less than a high school degree, spousal coverage contributes to the decline. Furthermore, among part-time workers, declines are explained entirely by reductions in eligibility for employer coverage—takeup among part-time workers actually increased among these workers (Farber and Levy, 2000). Thus, various mechanisms operate in the coverage decline within subgroups of workers.

Estimates of insurance coverage and trends differ based on the survey used, the time period, and the population of focus. The survey most commonly used for estimating health insurance coverage is the Current Population Survey (CPS), a household survey intended to collect monthly employment statistics. In March of each year, the CPS includes questions on health insurance coverage in the prior year. These questions have been changed periodically, creating difficulty in assessing trends over time. For example, in 1988, changes were made to questions and to the number of people asked about employer or union-sponsored coverage (Swartz 1997). Next, changes to question wording, sequencing, and sampling weights occurred in 1995, along with the addition of new questions (Swartz 1997). The question changes altered estimates of those who hold employment-based as opposed to individually purchased private insurance, affecting estimates beginning in 1994. For these reasons, some CPS trend estimates focus on subgroups for whom estimates were expected to be less affected (Kronick and Gilmer 1999) or on a limited time period (Acs 1995; Holahan et al., 1995; Holahan and Kim, 2000).

Another source, the CPS Benefits Supplement and Contingent Worker Supplements, analyzed by Farber and Levy (2000) and Currie and Yelowitz (1999), asks more detailed questions than the March CPS regarding coverage, employer offer of coverage, employee eligibility, and takeup. Although these supplements also differ over time, estimates of trends in coverage of adult workers by their own employer from these supplements are similar to estimates from the Survey of Income and Program Participation (SIPP). The comparability of results provides some indication that data in the supplements tracks with that of other surveys (Currie and Yelowitz, 1999). Finally, the Medical Expenditure Panel Survey (MEPS), analyzed by Cooper and Schone (1997) and Monheit and Schone (2000), collects detailed information on health insurance coverage, health care use, and expenditures, but the survey is conducted infrequently. The two most recent years for which data are available are 1996 and 1987 (through the MEPS' precursor, the National Medical Expenditure Survey).

Possible Explanations for Coverage Declines
A number of theories have been proposed to explain the decline in employment-based insurance coverage between the 1980s and 1990s. Trends in insurance coverage could reflect changes in labor patterns, in demographic characteristics of workers, crowdout, or changes in structural aspects of health insurance markets, such as increases in premium costs. The evidence points to structural aspects of health insurance coverage, rather than to changes in labor markets, as the more important factors in the declines in private coverage.

One set of possible explanations for the decline in coverage centers around changes in labor markets, such as industry shifts, use of part-time workers, and declines in unionization. These factors contribute partially to the decline. One study found that roughly a quarter (23%) of the decline in coverage for workers between 1988 and 1993 was due to the shift from manufacturing to service industries (10%), increased reliance on part-time workers (7%), and reduced unionization (6%) (Fronstin and Snider, 1996). Results from another analysis indicate that shifts in part-time and self-employed workers do not contribute to the decline in coverage, and that industry shifts explain up to 15 percent of the decline between 1980 and 1987 (Long and Rodgers, 1995).

Trends in wage levels are another important consideration in coverage rates, but also do not provide a full explanation.3 ACS (1995) finds that falling family incomes account for the bulk of the overall decline in insurance coverage between 1988 and 1991, but that a "secular" decline across all demographic, industry, and firm characteristics accounts for the decline in coverage among workers. Fronstin and Snider (1996) find that declines in real wages account for 23 percent of the decline in coverage between 1988 and 1993. However, if income trends were a primary explanation for coverage declines, one might expect to see a similar decline in pensions, a benefit similar in terms of value, tax preference, and risk pooling. Currie and Yelowitz (1999) make this comparison and show that unlike health insurance, pension benefits are more likely to be offered to and accepted by workers over time, particularly among workers with a college degree. In contrast, rates of health insurance coverage declined among college educated as well as less skilled workers. These results suggest that additional factors are contributing to coverage declines.

Another possibility is that expansions in public insurance programs, primarily the Medicaid program, have resulted in "crowd out" of privately insured individuals into public programs. While crowd out may be measured to answer several distinct questions, in the context of this paper, the relevant question is the extent to which Medicaid expansions contributed to declines in employer coverage. Generally, crowd out is a small to moderate factor in explaining declines in insurance coverage. Cutler and Gruber (1996) estimate that expansions of Medicaid coverage are associated with a decline in private insurance of 1.7 million persons, or 17 percent of the 9.9 million person decline in private insurance between 1987 and 1992. Currie and Yelowitz (1999) find that crowdout is a minor component of trends in employment-based insurance coverage but does not fully explain the decline.

page 1 | page 2

Fall 2000, Volume 1, Number 1
Table of Contents
Editor's Note
Features: Election 2000
Health Highlights
In Focus
Glossary of Health Care Terms

about us | links | contact us | subscribe | epihc