Article| Volume 18, ISSUE 1, P7-16, January 2008
• PDF [143 KB]PDF [143 KB]
• Top

# Women’s Health Insurance Coverage 1980–2005

• Author Footnotes
7 Sherry Glied, PhD, is Professor and Chair in the Department of Health Policy and Management at Columbia University’s Mailman School of Public Health. She is an economist whose research focuses on health insurance coverage and mental health policy.
8 Kathrine Jack, JD, is a recent graduate of Indiana University School of Law. She is currently employed as a special project attorney at the National Advocates for Pregnant Women. She wrote this paper while working as a research assistant at Columbia University.
9 Jason Rachlin, BA, MS, is currently working at the Bank of Montreal specializing in the risk management of their trading operations. When he wrote this paper, he was a research assistant at Columbia University.

### Background

In the fragmented US health insurance system, women’s health insurance coverage is an outcome both of changes in the availability of private and public health insurance and of changing patterns of labor force participation and household formation. Over the past 2 decades, women’s socioeconomic circumstances have changed and public policy around health insurance coverage for low-income women has also undergone substantial modification.

### Methods

This study examines the roles of these changes in circumstances and policy on the level and composition of women’s health insurance. Using the Census Bureau’s March Current Population Survey 1980–2005, the government’s principal source of nationally representative labor market and health insurance data, we examine how changes in marriage, full-time and part-time labor force participation, and public policy around coverage affected the level and source of women’s health insurance coverage over 3 periods: 1980–1987, 1988–1994, and 1995–2005.

### Results

Health insurance coverage rates have fallen for both women and men since 1980. What makes women different is that, in addition to the decline in coverage, the composition of health insurance coverage for women has also changed markedly. More women now obtain health insurance on their own, rather than as dependents, than did in 1980. A larger fraction of insured women are now enrolled in Medicaid than were in 1980. Women’s routes to coverage have changed as their social and economic circumstances have changed and as policy, especially Medicaid policy, has evolved.

### Conclusions

Women’s channels for obtaining health insurance coverage are more fragmented than those of men. The availability of multiple sources of coverage, and the possibility of moving amongst them, have not, however, insulated women from the overall declines in health insurance coverage caused by the rising cost of private health insurance.

## Changing Avenues to Coverage

### How women get coverage today

There are 4 main avenues through which working-age women (25–64) gain health insurance coverage (Figure 1). To give a sense of the importance of these avenues, we use the Current Population Surveys (CPS) data (described in detail below) to illustrate the rates of insurance coverage in 1980 and 2005 for different groups of working-age women (Table 1).
Table 1Women Age 25–64: Changing Patterns of Health Insurance by Work Status and Income, 1980–2005
198019871988199419952005
N53,642,20662,049,51763,014,72767,909,73368,691,81078,375,003
Insurance
Uninsured11.7%13.4%12.0%14.8%15.3%18.2%
Public coverage9.4%10.0%9.9%11.6%10.9%10.7%
ESI, from any source68.0%67.0%69.3%65.0%67.9%66.2%
ESI, own35.5%39.2%39.0%41.9%42.9%42.4%
ESI, dependent32.5%27.9%30.2%23.1%24.9%23.8%
Nongroup private insurance10.9%9.6%8.8%8.6%6.0%4.9%
Demographics
Full-time workers33.4%39.9%41.3%44.4%45.0%48.6%
Nonworkers34.5%29.6%28.6%26.0%25.2%26.4%
Married72.4%68.9%68.1%65.7%66.2%63.8%
Low income32.8%28.9%30.0%32.6%33.2%31.0%
Source: Analysis of CPS March Supplements.
Low-income = family incomes below 100% of poverty. A full-time worker worked ≥35 hours per week for ≥50 weeks in the last year. Part-time workers either worked <35 hours per week for any number of weeks or worked ≥35 hours per week for <50 weeks last year.
The most important avenue for coverage for women today is employment. About two thirds of all women obtain their health insurance coverage through employment (either their own or their spouse’s), and about two thirds of this group obtain it through their own jobs. Women are more likely than men to work for firms that offer coverage, mainly because they are more likely to work in large firms (
• Chollet D.
Employer-based health insurance in a changing workforce.
). Even when coverage is offered, however, some women may not be eligible for their employers’ programs because many women work only part time. In 2004, women constituted 46.7% of all workers but only 41.4% of all full-time workers (authors’ tabulations of March 2005 CPS). Some women do not enroll in their employers’ health insurance plan even when coverage is offered and they are eligible. Workers with an employed spouse are more likely to decline coverage from their own employer (
• Buchmueller T.C.
Marital status, spousal coverage, and the gender gap in employer-sponsored health insurance.
). Low-income workers are also more likely to decline coverage (
• Blumberg L.
• Nichols L.
The health status of workers who decline employer-sponsored insurance.
). Women are overrepresented in both categories.
Many women obtain coverage through spousal or domestic partner’s employers, the second most common source of coverage for working age women. In 1993, 78% of working married men were offered employer-sponsored health insurance coverage for their dependents (
• Buchmueller T.C.
Marital status, spousal coverage, and the gender gap in employer-sponsored health insurance.
). Women are twice as likely as men to be insured through spousal coverage (
• Lambrew J.M.
Diagnosing disparities in health insurance for women: A prescription for change.
).
Although more firms now offer same and opposite sex domestic partner health insurance coverage, enrollment in such program is very low (
• Costello A.C.
Domestic partner benefits in the workplace.
).
A minority of working-age women is eligible for public health insurance. Because Medicare eligibility for those <65 years old is disability based, most working-age women with public coverage are covered through Medicaid. Eligibility for Medicaid is based on both income and categorical status. Low-income adults may be categorically eligible for Medicaid because of a disability or because they have dependent children. Women are more likely to qualify under these criteria because they have lower incomes and are more likely to live with dependent children than are men. Eligibility for low-income pregnant women offers some women a route to coverage not available to men.
Finally, a few women purchase health insurance coverage in the individual insurance market. In 2000, 4.9% of women obtained coverage in this way (authors’ tabulation of CPS data). Women are slightly more likely to purchase individual insurance than men, but women who do so are more likely to be single, older, and have lower incomes than men who purchase individual insurance (
• Lambrew J.M.
Diagnosing disparities in health insurance for women: A prescription for change.
).

### Changing economic and social factors affecting health insurance coverage: 1980–2005

The 4 routes through which women obtain coverage were the same in 2005 as in 1980, but significant social and economic changes have occurred that have altered the relative importance of each avenue. Table 1 describes the changes in economic and social circumstances of women during this period.
The biggest change over this period has been women’s growing attachment to the labor force. Women’s labor force participation rates rose continually throughout this period, especially between 1980 and 1990. The female labor force participation rate increased by 8 percentage points between 1980 and 2005 (authors’ tabulations of the CPS data). Participation increased substantially for all but women near retirement age, and increased the most for young women (
• Blau F.
Trends in the well-being of American women, 1970–1995.
). Full-time work increased over this period, and part-time work declined. Other aspects of women’s labor force participation, not shown in Table 1, reinforce these patterns. Women are increasingly likely to be working in jobs similar to those held by men. Occupational segregation declined substantially over this period (
• Blau F.
Trends in the well-being of American women, 1970–1995.
). At the same time, earnings for working-age women also increased relative to those of men. Women’s earnings increased at least in part because of their growing labor market experience and increasing representation in traditionally male fields. The gender wage gap narrowed rapidly during the 1980s, and continued to narrow during the 1990s (
• Blau F.
Trends in the well-being of American women, 1970–1995.
). Because working women can obtain employer-sponsored insurance, these developments would be expected to lead to increased health insurance coverage.
Increases in women’s labor force participation are contemporaneous with deferment of marriage and childbearing and an increasing rate of divorce. The proportion of single women ages 25–64 rose >8%, with most of the increase occurring during the 1980s. The fraction of women currently married has dropped by 10% (
• Blau F.
Trends in the well-being of American women, 1970–1995.
). The proportion of women who were not currently married but who were previously married increased slightly over this period from 20% to 21.4% (authors’ tabulation of CPS data).
Despite these substantial changes in employment and marriage, women’s poverty status changed relatively little over this period. Slightly more women had incomes >100% of the federal poverty level (FPL) in 2005 than in 1980.

### Changes in public policy and in the health insurance market between 1980 and 2005

The changes in economic and social circumstances described have interacted with changes in policy and in private health insurance availability to generate trends in health insurance outcomes.
The availability of public insurance coverage for women changed over this period for 2 reasons. First, expansions in eligibility for Medicaid led to increases in the number of women eligible for public coverage. Changes in Medicaid eligibility occurred through expanded categorical eligibility and increases in income eligibility levels. For most of this period, women eligible for Aid for Families with Dependent Children (AFDC) were automatically eligible for Medicaid. Eligibility rules for AFDC were income and asset based, but varied by state. In 1980, AFDC was limited to single-parent households, with a few exceptions (
• Winkler A.E.
AFDC-UP, two-parent families, and the Family Support Act of 1988: Evidence from the 1990 CPS and the 1987 NSFH.
). During the 1980s, many states introduced AFDC-UP, a program through which low-income, married couple families where the household head becomes unemployed could gain AFDC temporarily. The federal government required all states to offer AFDC-UP to such families by 1991. This program affected about 5% of AFDC families in 1990 (
• Currie J.
• Gruber J.
Saving babies: The efficacy and cost of recent expansions of Medicaid eligibility for pregnant women.
).
Beginning in 1986, the federal government passed a series of laws encouraging states to expand Medicaid eligibility to low-income children and pregnant women, regardless of their eligibility for AFDC. All states were required to raise the income eligibility threshold for pregnant women to 133% of the FPL eligible for Medicaid by 1990 (
National Governors’ Association
State coverage of pregnant women and children.
). Most states raised the income limits beyond the federal minimum. By 2002, a majority of states deemed a pregnant woman eligible for Medicaid if she has an income <185% of the FPL. The effect of these expansions was substantial. The percentage of women 15–44 who would be eligible for Medicaid if they were pregnant rose from 12.4% in 1979 to 43.3% in 1991. Just >11% of women in this age range are pregnant at some point during the year, suggesting that this expansion affected about 4% of all women 15–44 (
• Currie J.
• Gruber J.
Saving babies: The efficacy and cost of recent expansions of Medicaid eligibility for pregnant women.
). For more detailed information about the Medicaid expansions, see
• Currie J.
• Grogger J.
Medicaid expansions and welfare contractions: Offsetting effects on prenatal care and infant health?.
and
• Currie J.
• Gruber J.
Saving babies: The efficacy and cost of recent expansions of Medicaid eligibility for pregnant women.
.
Second, offsetting these expansions, the shift from AFDC to the Temporary Assistance for Needy Families program in 1997 severed the connection between cash assistance and health insurance eligibility for many women. Changes in the connections between Medicaid eligibility and welfare in the 1990s have contributed to a reduction in the participation rate for Medicaid among eligible women (
• Pati S.
• Romero D.
• Chavkin W.
Changes in use of health insurance and food assistance programs in medically underserved communities in an era of welfare reform: An urban study.
).
There have also been substantial changes in private insurance coverage over this period that have affected the ability of women to obtain coverage through their own jobs or those of their spouses. Most of the period examined in this paper saw very sharp and rapid increases in health care costs.
From 1993 to 1999, health care costs grew slower than the overall economy (
Centers for Medicare and Medicaid Studies
National health expenditures.
).
These increases led to declines in the prevalence of employer-sponsored health insurance among men and women (
• Glied S.
• Stabile M.
Explaining the decline in insurance coverage among young men.
,
• Kronick R.
• Gilmer T.
Explaining the decline in health insurance coverage: 1979–1995.
).
The rate at which employers offered coverage was relatively stable over most of this period (
• Cooper P.F.
• Schone B.S.
More offers, fewer takers for employment-based health insurance: 1987 and 1996.
,
• Cutler D.M.
Employee costs and the decline in health insurance coverage.
). Required annual employee contributions for coverage, however, rose rapidly between 1988 and 1993, increasing from an average of about $125 to >$450 (in constant dollars) for single coverage, and doubling from $814 to >$1,650 for family coverage (
• Cutler D.M.
Employee costs and the decline in health insurance coverage.
). Required contribution levels jumped again between 2001 and 2004, increasing by about 25%–30% (inflation adjusted) for single and family coverage (
Kaiser Family Foundation Trends and indicators in the changing health care marketplace.
). These increases in required contributions for coverage led to substantial declines in the take-up rate for coverage, generating an overall decline in employer-sponsored insurance (
• Cutler D.M.
Employee costs and the decline in health insurance coverage.
).

## Data and Methods

### Data

We use data on women 25–64 from the March Demographic Supplements of the CPS, conducted by the Census Bureau, for 1980, 1987, 1988, 1994, 1995, and 2005 (
US Census Bureau
Current population survey design and methodology [technical paper 63RV].
).
We use a public release of these data processed by the Unicon Research Corporation.
The CPS is a stratified sample of approximately 50,000 households each month examining labor force participation of each member of the survey households. When weighted using the Census Bureau provided weights (as we do throughout these analyses), it produces nationally representative samples. The March Supplement includes additional demographic information on each individual as well as information on health insurance coverage.
We focus on women 25–64. The lower age bound is set at 25 because these women are no longer eligible for dependent coverage from their parents and are likely to have completed college. Women age ≥65 are eligible for Medicare.
The March CPS began asking respondents about health insurance coverage in 1980. Respondents are asked about coverage through Medicare, Medicaid, CHAMPUS, employer-sponsored insurance, and individually purchased coverage. In principle, these questions refer to coverage obtained at any time during the preceding calendar year.
Although CPS data in principle refers to the year before they are released, we use the release year to identify year.
Those with private insurance are asked whether it was obtained through their own employer, as a dependent, or through individual purchase.

### Methods

Most of the variables used in our analyses are drawn directly from the CPS. We define full-time workers as those who report working full time (>35 hours per work) for a full year (>50 weeks), whether they are employed or self-employed, and nonworkers as those who report no work in the prior year. We define part-time workers as those who are neither full-time workers nor nonworkers according to these definitions. We define married as currently married, regardless of whether the spouse is present or absent.
In the regression analyses used in the decomposition, we also include measures of education, race, and ethnicity. We construct education measures (less than high school, high school, at least some college) using years of education completed. We define Hispanic as any person of Hispanic origin and nonwhite as any person reporting ethnicity other than white.
We classify adult women receiving employer-sponsored coverage as dependents as obtaining “spousal” employer-sponsored insurance. Women who receive insurance through a domestic partner are included in the “spousal” employer-sponsored insurance category. Fewer than 1% of women ages 25–64 received employer-sponsored insurance as a dependent and are not married during each of the years examined.
In most years of the CPS, lack of insurance is computed as a residual. The CPS questions have changed periodically since 1980, with major changes occurring in 1988 and in 1995 (see
• Swartz K.
Changes in the 1995 current population survey and estimates of health insurance coverage.
for a discussion of these changes).
The changes in 1988 and 1995 both led to an increase in the estimated number of people with private insurance. Weighting changes after the 1990 census were adjusted for using a weight bridge file in the 1993 CPS.
These changes make it difficult to make cross-year comparisons for many aspects of insurance coverage (
• Levit K.
• Olin G.L.
• Letsch S.W.
Americans’ health insurance coverage, 1980–1991.
). As a consequence, we have examined changes between these endpoints for most of our analyses. Beginning in 2001, the CPS also added an uninsurance “verification” question, but we have calculated our results for years after 2001 using the earlier (unverified) insurance question to create a consistent time series.
We analyzed women’s insurance coverage using an accounting decomposition based on the routes to coverage described in Figure 1. The possibility of participating in each of the 4 coverage types (employer sponsored through own employment, employer sponsored through spouse’s employment, individual coverage, and public coverage) depends on a woman’s circumstances. Employer-sponsored health insurance provided by one’s own employer can only be obtained by full- or part-time workers. Only married women can obtain spousal employer-sponsored health insurance. Any women can obtain public insurance and individual insurance. For public coverage, however, low-income and higher income women face different eligibility. We operationalize this by comparing women with incomes above and below 100% FPL.
For each period, we computed the size of the population eligible for each coverage route and the percentage with that particular insurance type for each year examined. The first row in Figure 1 represents the routes. The arrows represent the propensity for the type of coverage in the box below. We then computed the share of the change in overall insurance coverage for all women associated with the changes in the size of the population eligible for each route to coverage and the propensity for women with these circumstances to receive that form of coverage. The sum of the population in each avenue multiplied by the propensity for coverage within that group equals the total health insurance coverage among working-age women. Note that this type of analysis does not account for overlap among the avenues. For example, a woman could be both a full-time worker and married.
We repeat the analysis using a standard labor economics technique, the Blinder–Oaxaca decomposition, to confirm that our results are stable after considering overlaps among population categories and controlling for other changes (
• Reilly K.T.
• Wirjanto T.S.
Does more mean less? The male/female wage gap and the proportion of females at the establishment level.
). We performed this decomposition separately for overall coverage, private coverage, and public coverage. Similar regression-based decomposition analyses of insurance coverage using the CPS data have been performed in prior research (
• Acs G.
Explaining trends in health insurance coverage between 1988 and 1991.
,
• Long S.H.
• Rodgers J.
Do shifts towards service industries, part-time work, and self-employment explain the rising uninsured rate?.
,
• Fronstin P.
• Snider S.C.
An examination of the decline in employment-based health insurance between 1988 and 1993.
). We use the base year of each pairwise comparison (1980 for the 1980–1987 comparison; 1988 for the 1988–1994 comparison; 1995 for the 1995–2005 comparison) as the point of reference. The basic formula for the Oacaxa decomposition is as follows:
$HI87−HI80=X87*(B87−B80)+B80*(X87−X80)$

where HI measures health insurance, X is a regressor, and B is a vector of coefficients that give the probabilities of obtaining insurance for each X. In the analyses that follow, we include 4 main X vectors: full-time work status, labor force nonparticipation, marital status, and low income, defined as <100% of FPL. We control for education, region, race, and ethnicity. Regression results for each period of analysis and insurance type are shown in Appendix 1 Table, Appendix 2 Table, Appendix 3 Table.
Appendix 1 TableAny Insurance Coverage: Linear Regression Coefficients
198019871988199419952005
N53,642,20662,049,51763,014,72767,909,73368,691,81078,375,003
Full-time workers0.0414* (0.0001)0.0727* (0.0001)0.0537* (0.0001)0.0608* (0.0001)0.0662* (0.0001)0.0694* (0.0001)
Nonparticipation in labor force−0.0168* (0.0001)0.0105* (0.0001)0.0424* (0.0001)0.0515* (0.0001)0.0337* (0.0001)0.0270* (0.0001)
Married now0.0491* (0.0001)0.0758* (0.0001)0.0577* (0.0001)0.0450* (0.0001)0.0705* (0.0001)0.0694* (0.0001)
<100% FPL−0.1755* (0.0001)−0.1645* (0.0001)−0.2149* (0.0001)−0.1816* (0.0001)−0.1447* (0.0001)−0.1817* (0.0001)
Northeast region0.0540* (0.0001)0.0456* (0.0001)0.0538* (0.0001)0.0332* (0.0001)0.0264* (0.0001)0.0208* (0.0001)
North central region0.0514* (0.0001)0.0354* (0.0001)0.0425* (0.0001)0.0327* (0.0001)0.0329* (0.0001)0.0145* (0.0001)
South0.0081* (0.0001)−0.0063* (0.0001)−0.0009* (0.0001)−0.0181* (0.0001)−0.0143* (0.0001)−0.0319* (0.0001)
Nonwhite−0.0033* (0.0001)−0.0232* (0.0001)−0.0173* (0.0001)−0.0141* (0.0001)−0.0229* (0.0001)−0.0354* (0.0001)
Hispanic ethnicity−0.0660* (0.0002)−0.1051* (0.0002)−0.1214* (0.0002)−0.1109* (0.0002)−0.1464* (0.0001)−0.1614* (0.0001)
College graduate0.0338* (0.0001)0.0506* (0.0001)0.0523* (0.0001)0.0582* (0.0001)0.0604* (0.0001)0.0816* (0.0001)
Intercept0.8267* (0.0002)0.7771* (0.0001)0.8047* (0.0001)0.7860* (0.0001)0.7634* (0.0001)0.7446* (0.0001)
r20.0740.090.1160.0880.0880.111
Standard errors in parentheses with p < .05 = ∼, p < .01 = *.
Abbreviation: FPL, federal poverty level.
Appendix 2 TablePrivate Insurance Coverage: Linear Regression Coefficients
198019871988199419952005
N53,642,20662,049,51763,014,72767,909,73368,691,81078,375,003
Full-time workers0.0824* (0.0001)0.1109* (0.0001)0.1024* (0.0001)0.1163* (0.0001)0.1326* (0.0001)0.1302* (0.0001)
Nonparticipation in labor force−0.1336* (0.0001)−0.1282* (0.0001)−0.0797* (0.0001)−0.0895* (0.0001)−0.1112* (0.0001)−0.1255* (0.0001)
Married now0.1498* (0.0001)0.1597* (0.0001)0.1354* (0.0001)0.1306* (0.0001)0.1726* (0.0001)0.1623* (0.0001)
<100% FPL−0.3870* (0.0002)−0.4154* (0.0002)−0.4432* (0.0001)−0.4333* (0.0001)−0.3180* (0.0001)−0.2896* (0.0001)
Northeast region0.0617* (0.0001)0.0582* (0.0001)0.0583* (0.0001)0.040* (0.0001)0.0427* (0.0001)0.0118* (0.0001)
North central region0.0699* (0.0001)0.050* (0.0001)0.0511* (0.0001)0.0344* (0.0001)0.0446* (0.0001)0.0172* (0.0001)
South0.0207* (0.0001)0.0141* (0.0001)0.0081* (0.0001)−0.0112* (0.0001)0.0012* (0.0001)−0.0230* (0.0001)
Nonwhite−0.0805* (0.0001)−0.0742* (0.0001)−0.0881* (0.0001)−0.0640* (0.0001)−0.0910* (0.0001)−0.0678* (0.0001)
Hispanic ethnicity−0.0872* (0.0002)−0.1108* (0.0002)−0.1391* (0.0002)−0.1272* (0.0002)−0.1553* (0.0002)−0.1502* (0.0001)
College graduate0.0537* (0.0001)0.0749* (0.0001)0.0794* (0.0001)0.0948* (0.0001)0.0955* (0.0001)0.1341* (0.0001)
Intercept0.7079* (0.0002)0.6620* (0.0002)0.6942* (0.0002)0.6527* (0.0002)0.6034* (0.0002)0.5781* (0.0002)
r20.2760.3160.3390.3380.2970.276
Standard errors in parentheses with p < .05 = ∼, P < .01 = *.
Abbreviation: FPL, federal poverty level.
Appendix 3 TablePublic Insurance Coverage: Linear Regression Coefficients
198019871988199419952005
N53,642,20662,049,51763,014,72767,909,73368,691,81078,375,003
Full-time workers−0.0409* (0.0001)−0.0382* (0.0001)−0.0488* (0.0001)−0.0554* (0.0001)−0.0665* (0.0001)−0.0608* (0.0001)
Nonparticipation in labor force0.1168* (0.0001)0.1386* (0.0001)0.1221* (0.0001)0.1410* (0.0001)0.1450* (0.0001)0.1526* (0.0001)
Married now−0.1007* (0.0001)−0.0839* (0.0001)−0.0778* (0.0001)−0.0856* (0.0001)−0.1022* (0.0001)−0.0930* (0.0001)
<100% FPL0.2115* (0.0001)0.2509* (0.0001)0.2283* (0.0001)0.2517* (0.0001)0.1733* (0.0001)0.1079* (0.0001)
Northeast region−0.0077* (0.0001)−0.0126* (0.0001)−0.0045* (0.0001)−0.0068* (0.0001)−0.0163* (0.0001)0.0091* (0.0001)
North central region−0.0185* (0.0001)−0.0146* (0.0001)−0.0086* (0.0001)−0.0017* (0.0001)−0.0117* (0.0001)−0.0026* (0.0001)
South−0.0127* (0.0001)−0.0205* (0.0001)−0.0090* (0.0001)−0.0069* (0.0001)−0.0155* (0.0001)−0.0089* (0.0001)
Nonwhite0.0772* (0.0001)0.0510* (0.0001)0.0708* (0.0001)0.0498* (0.0001)0.0681* (0.0001)0.0324* (0.0001)
Hispanic ethnicity0.0213* (0.0002)0.0057* (0.0001)0.0176* (0.0001)0.0163* (0.0001)0.0089* (0.0001)−0.0112* (0.0001)
College graduate−0.020* (0.0001)−0.0244* (0.0001)−0.0271* (0.0001)−0.0366* (0.0001)−0.0351* (0.0001)−0.0524* (0.0001)
Intercept0.1189* (0.0001)0.1151* (0.0001)0.1105* (0.0001)0.1332* (0.0001)0.160* (0.0001)0.1665* (0.0001)
r20.2060.2320.2270.2610.2210.171
Standard errors in parentheses with p < .05 = ∼, p < 0.01 = *.
Abbreviation: FPL, federal poverty level.

## Results

Panel A of Table 2A, Table 2B, Table 3A, Table 3B, Table 4A, Table 4B provides the results of the accounting decomposition for the 1980–1987, 1988–1994, and 1995–2005 periods, respectively. In each table, the first column gives the percentage of all women obtaining coverage through that route in the first year of analysis. The second column, ΔB, indicates how changes in the propensity to obtain coverage conditional on being in that category affected the probability of gaining coverage over the period. The third column, ΔX, indicates how changes in the percentage of women in this category affected the probability of gaining this type of coverage over this period. We report a zero in this column for individual coverage when the eligible population group includes all women. The final column, ΔHI, indicates how changes in that category affected overall health insurance coverage among women.
Table 2AAccounting Decomposition of Health Insurance Changes 1980–1987, Working-Age Women
% of Women in this Category in 1980ΔB (%)ΔX* (%)ΔHI (%)
Any health insurance coverage88.32−1.710.00−1.71
Any Private health insurance78.92−2.350.00−2.35
Health insurance from own employer among full time workers24.290.314.705.01
Health insurance from own employer among part time workers11.18−0.78−0.56−1.34
Health insurance from spouse’s employer32.50−3.08−1.56−4.64
Nongroup private health insurance10.91−1.360.00−1.36
Any Public health insurance9.410.630.000.63
Public coverage among low income4.900.54−0.100.44
Public coverage among higher income4.510.170.010.19
Table 2BBlinder–Oaxaca Decomposition for Health Insurance Changes 1980–1987, Working-Age Women
Total Changes Due to Coefficients (Propensity) (%)Change in Means (Size of the Population in Each Avenue) (%)Total Change in Insurance (%)
Any coverage−1.870.17−1.71
Full-time workers1.250.27
Nonworkers0.810.08
Married women1.84−0.17
Low-income women0.140.05
Other characteristics and the constant−4.28−0.06
Private coverage−2.960.62−2.35
Full-time workers1.150.53
Nonworkers0.160.65
Married women0.68−0.52
Low-income women−0.360.11
Other characteristics and coefficient of the constant−4.59−0.14
Public coverage1.12−0.460.63
Full-time workers0.11−0.26
Nonworkers0.65−0.57
Married women1.160.35
Low-income women0.51−0.06
Other characteristics and coefficient of the constant−1.290.08
Low income is defined as <100% of the federal poverty level. Full-time workers defined as full-year, full-time employment.
Table 3AAccounting Decomposition of Health Insurance Changes 1988–1994, Working-Age Women
% of Women in this Category in 1988ΔB (%)ΔX* (%)ΔHI (%)
Any health insurance coverage87.96−2.770.00−2.77
Any private health insurance78.04−4.490.00−4.49
Health insurance from own employer among full-time workers29.11−0.342.241.91
Health insurance from own employer among part-time workers8.410.66−0.140.51
Health insurance from spouse’s employer30.24−6.11−1.04−7.15
Nongroup private health insurance8.76−0.090.00−0.09
Any public health insurance9.921.720.001.72
Public coverage among low income5.570.660.771.43
Public coverage among higher income4.350.39−0.100.29
Table 3BBlinder–Oaxaca Decomposition for Health Insurance Changes 1988–1994, Working-Age Women
Total Changes Due to Coefficients (Propensity) (%)Change in Means (Size of the Population in Each Avenue) (%)Total Change in Insurance (%)
Any coverage−2.43−0.33−2.76
Full-time workers0.320.17
Nonworkers0.24−0.11
Married women−0.83−0.14
Low-income women0.53−0.41
Other characteristics and the constant−2.170.16
Private coverage−4.10−0.40−4.49
Full-time workers0.620.33
Nonworkers−0.250.21
Married women−0.32−0.32
Low-income women0.16−0.85
Other characteristics and coefficient of the constant−4.310.24
Public coverage1.670.061.72
Full-time workers−0.29−0.15
Nonworkers0.49−0.33
Married women−0.510.18
Low-income women0.370.44
Other characteristics and coefficient of the constant1.62−0.08
Low income is defined as <100% of the federal poverty level. Full-time workers defined as full-year, full-time employment.
Table 4AAccounting Decomposition of Health Insurance Changes 1995–2005, Working-Age Women
% of Women in this Category in 1995ΔB (%)ΔX* (%)ΔHI (%)
Any health insurance coverage84.71−2.880.00−2.88
Any Private health insurance73.84−2.670.00−2.67
Health insurance from own employer among full-time workers31.48−1.592.530.94
Health insurance from own employer among part-time workers9.33−0.07−1.49−1.56
Health insurance from spouse’s employer24.74−0.35−0.89−1.24
Nongroup private health insurance5.96−1.020.00−1.02
Any public health insurance10.88−0.220.00−0.22
Public coverage among low income8.40−0.13−0.57−0.70
Public coverage among higher income2.480.390.080.48
Table 4BBlinder–Oaxaca decomposition for Health Insurance Changes 1995–2005, Working-Age Women
Total Changes Due to Coefficients (Propensity) (%)Change in Means (Size of the Population in Each Avenue) (%)Total Change in Insurance (%)
Any coverage−2.88−0.01−2.89
Full-time workers0.160.24
Nonworkers−0.180.04
Married women−0.07−0.17
Low-income women−0.560.03
Other characteristics and the constant−1.69−0.15
Private coverage−2.55−0.12−2.67
Full-time workers−0.120.48
Nonworkers−0.38−0.13
Married women−0.66−0.41
Low-income women0.430.07
Other characteristics and coefficient of the constant−1.83−0.13
Public coverage−0.320.10−0.21
Full-time workers0.28−0.24
Nonworkers0.200.17
Married women0.590.24
Low-income women−0.99−0.04
Other characteristics and coefficient of the constant−0.39−0.03
Low income is defined as <100% of the federal poverty level. Full-time workers defined as full-year, full-time employment.
Panel B of Table 2A, Table 2B, Table 3A, Table 3B, Table 4A, Table 4B displays the results of the corresponding Blinder–Oaxaca decompositions for overall, private, and public insurance for the 3 periods. The first row shows the change in insurance coverage for the period. The following rows display the change in health insurance attributable to the changing propensity and to the size of the population in each route to coverage.
We note that t-tests of the difference in mean rates of coverage within each period show that the overall declines in coverage documented here were statistically significant in each period.
In the first period of analysis, 1980–1987, overall health insurance coverage declined as private coverage fell >2 percentage points and public coverage grew only slightly. This period saw a substantial increase in full-time work among women (4.7%) and the accounting decomposition indicates that this increase would have been expected to lead to an increase in coverage. This potential increase in coverage was almost fully offset, however, by a decline in the proportion of women who were married (−1.56%) and by a substantial decline in the propensity of married women to obtain spousal coverage (−3.08%). Overall, the share of women with employer-sponsored coverage fell slightly, but the composition changed dramatically. Many more women held coverage in their own name through their own full-time jobs and many fewer were covered through a spouse.
The Blinder–Oaxaca decomposition similarly shows that a reduction in the propensity to obtain private coverage (−2.96%) was the principal factor in the decline in insurance over this period. Controlling for changes in income and education in this decomposition reduces the estimated positive effect of increased full-time work (0.53%) (compared to with the accounting decomposition), but even after controlling for other characteristics, we find that changes in women’s circumstances would have been expected to lead to a substantial increase in private coverage over this period (0.62%). However, the declining propensity to obtain such coverage more than cancelled out this effect.
Both analyses suggest that a slight decline in the share of women who had low incomes contributed to reduced public coverage, but was offset by an increased propensity for low-income women to obtain public coverage as expanded eligibility for public programs took effect. Increases in public coverage offset about one quarter of the decline in private coverage.
Table 3A, Table 3B repeats these analyses for the second time period, 1988–1994. Health insurance coverage for women declined sharply and statistically significantly in this period, with private coverage falling almost 4.5 percentage points. About 1.7 percentage points of this decline was offset by an increase in public coverage.
As the accounting decomposition shows, there was relatively little change in women’s socioeconomic circumstances in this period. Instead, the decline in private coverage was largely driven by a steep fall in spousal coverage among married women (−6.11%). Because of this sharp decline, and a simultaneous slight increase in full-time work (2.24%), by the end of this period, women were more likely to obtain coverage as full-time workers holding employer-sponsored coverage in their own name than through a spouse. The Blinder–Oaxaca decomposition similarly shows that changes in the means contributed little to coverage changes in this period, and changes in coefficients drove most of the change in coverage.
Public coverage increased during this period both because the share of women who were low income grew (0.77%) and because the propensity for low-income women to obtain public coverage again increased (0.66%). As in the prior period, expansions in eligibility led to increases in the propensity to obtain coverage.
Table 4A, Table 4B provides the results for 1995–2005. Overall health insurance coverage for women fell again, but during this period, a slight decline in public coverage (−0.22%) exacerbated, rather than offset, the decline in private coverage (−2.67%). The accounting decomposition shows that increases in full-time work (2.53%) in the buoyant economy of the late 1990s led to increases in the propensity for women to obtain their own employer-sponsored insurance coverage through full-time employment (0.94%), but these were offset by declines in the fraction of women holding coverage through part-time employment. At the same time, and in marked contrast to prior periods, married women were only slightly less likely to obtain spousal coverage at the end of this period than they had been at the beginning (−1.24%). The fraction of working-age women who were married continued to decline over this period, but the propensity of married women to obtain coverage through spousal employment declined only slightly. The Blinder–Oaxaca decomposition confirms these findings, showing that changes in the propensity to obtain coverage (−2.55%) rather than population changes (−0.12%) generated the decline in private coverage.
Declines in public coverage over this period occurred both because there were fewer low-income women who might qualify for public coverage and because those who were low income were slightly less likely to obtain this type of coverage. In the accounting decomposition, the decline in the share of women who were low income explains most of the decline in public coverage. In the Blinder–Oaxaca decomposition, controlling for other characteristics reduces the contribution of improved living circumstances. This analysis suggests that, after taking all the changes in women’s circumstances into account, the propensity for women to obtain public coverage has declined (−0.99%). This pattern is consistent with the changes in eligibility associated with welfare reform.

## Discussion

Women’s health insurance coverage has declined over the last 2 decades. The most important contributor to this decline for women—as for men—has been a declining propensity to obtain private coverage, conditional on socioeconomic circumstances. For women, however, the sharpest declines in propensities to obtain coverage have been for married women obtaining spousal coverage. The fraction of married women obtaining such coverage declined steeply between 1980 and 1987, and particularly between 1988 and 1994, while employee contributions for family coverage rose rapidly, before stabilizing in the 1995–2005 period.
A second important factor changing the composition of women’s coverage over this period has been the legislated expansion in the availability of public coverage. Increases in public coverage before 1995 offset a portion of the steep declines in private coverage for women, but a lack of further expansions in public coverage contributed to reductions in coverage after the mid 1990s.
Throughout this period, the effect on insurance of increases in full-time work among women has more than offset the effect of declining marriage rates. By 2005, women were about 78% more likely to be obtaining private coverage through their own full- or part-time jobs as through their spouses’ jobs. This change has been a consequence of both increasing full-time work and employer policy changes. Over the past 2 decades, employers have increasingly curtailed coverage or increased premium shares for dependent coverage, reducing the propensity for married women to obtain coverage through their spouses. The reduced importance of this mode of coverage means that fewer women are at risk from such changes in the future—and that the savings to employers from these strategies are likely to be much smaller in the future.
The composition of coverage for women has changed along with women’s changing circumstances. Women’s channels for obtaining health insurance coverage are more fragmented than those of men, who tend to get coverage mainly through their own employers. The availability of multiple sources of coverage, and the possibility of moving among them, have not, however, insulated women from the overall changes in health insurance experienced by men. As holds true for men, declines in the propensity of workers to obtain coverage, for themselves or their spouses, through employment, has been the major factor leading to declines in coverage since 1980 (
• Glied S.
• Stabile M.
Explaining the decline in insurance coverage among young men.
).
Changing propensity to obtain coverage in the private insurance markets is primarily a function of health care costs, whereas changing propensity in the public market reflects changing public policy. Overall, the results reported herein suggest that rising health care costs are the greatest factor influencing women’s health insurance coverage through declining private coverage. This trend is not unique for working-age women. However, other factors unique to working-age women had varied impact on insurance coverage. The decline in coverage was at times offset by increasingly generous public insurance programs. Increases in women’s full-time labor force participation were the biggest factor that expanded private coverage and helped women to buck the overall trend.
The results of these analyses suggest that improvements in the economy, alone, will not be enough to substantially reduce uninsurance rates among women in the future. Higher health care costs have consistently outpaced the effects of rising full-time work. As the cost of coverage rises relative to incomes, policy makers will need to assist workers to help them afford private insurance coverage. Alternatively, they will need to expand public programs to cover not only low-income children and pregnant women, but other low-income adults.

## References

• Acs G.
Explaining trends in health insurance coverage between 1988 and 1991.
Inquiry. 1995; 32: 102-110
• Blau F.
Trends in the well-being of American women, 1970–1995.
Journal of Economic Literature. 1998; 36: 112-165
• Blumberg L.
• Nichols L.
The health status of workers who decline employer-sponsored insurance.
Health Affairs (Millwood). 2001; 36: 180-187
• Buchmueller T.C.
Marital status, spousal coverage, and the gender gap in employer-sponsored health insurance.
Inquiry. 1996; 33: 308-316
• Chollet D.
Employer-based health insurance in a changing workforce.
Health Affairs. 1994; 13: 315-326
• Centers for Medicare and Medicaid Studies
National health expenditures.
(Accessed October 7, 2002)
• Cooper P.F.
• Schone B.S.
More offers, fewer takers for employment-based health insurance: 1987 and 1996.
Health Affairs. 1997; 16: 142-149
• Costello A.C.
Domestic partner benefits in the workplace.
Journal of Pension Benefits. 2002; 10: 10-14
• Currie J.
• Grogger J.
Medicaid expansions and welfare contractions: Offsetting effects on prenatal care and infant health?.
Journal of Health Economics. 2002; 21: 313-335
• Currie J.
• Gruber J.
Saving babies: The efficacy and cost of recent expansions of Medicaid eligibility for pregnant women.
The Journal of Political Economy. 1996; 104: 1263-1296
• Cutler D.M.
Employee costs and the decline in health insurance coverage.
Frontiers in Health Policy Research. 2003; 6: 27-54
• Fronstin P.
• Snider S.C.
An examination of the decline in employment-based health insurance between 1988 and 1993.
Inquiry. 1996; 33: 317-325
• Glied S.
• Stabile M.
Explaining the decline in insurance coverage among young men.
Inquiry. 2000; 37: 295-303
1. Kaiser Family Foundation Trends and indicators in the changing health care marketplace.
(Accessed July 10, 2007)
• Kronick R.
• Gilmer T.
Explaining the decline in health insurance coverage: 1979–1995.
Health Affairs. 1999; 18: 30-47
• Lambrew J.M.
Diagnosing disparities in health insurance for women: A prescription for change.
(p. 35) Commonwealth Fund, New York2001
• Levit K.
• Olin G.L.
• Letsch S.W.
Americans’ health insurance coverage, 1980–1991.
Health Care Financing Review. 1992; 14: 31
• Long S.H.
• Rodgers J.
Do shifts towards service industries, part-time work, and self-employment explain the rising uninsured rate?.
Inquiry. 1995; 32: 111-116
• National Governors’ Association
State coverage of pregnant women and children.
Author, Washington, DC1991
• Pati S.
• Romero D.
• Chavkin W.
Changes in use of health insurance and food assistance programs in medically underserved communities in an era of welfare reform: An urban study.
American Journal of Public Health. 2002; 92: 1441-1445
• Reilly K.T.
• Wirjanto T.S.
Does more mean less?.
Canadian Journal of Economics. 1999; 32: 906-929
• Swartz K.
Changes in the 1995 current population survey and estimates of health insurance coverage.
Inquiry. 1997; 34: 70-79
• US Census Bureau
Current population survey design and methodology [technical paper 63RV].
www.census.gov
Date: 2002
(Accessed November 4, 2007)
• Winkler A.E.
AFDC-UP, two-parent families, and the Family Support Act of 1988: Evidence from the 1990 CPS and the 1987 NSFH.
([discussion paper no. 1013-93]) Institute for Research on Poverty, Madison, WI1993 (August)