WIDOWHOOD DOESN’T CAUSE COVERAGE LOSS
Health care needs tend to increase with age. Thus the population approaching retirement age, sometimes called the “near elderly,” is especially vulnerable if they experience lapses in heath insurance coverage. A study by David Weir and Robert Willis, funded by the Economic Research Initiative on the Uninsured (ERIU), investigates the channels through which near-elderly women gain and lose health insurance coverage, focusing on widowhood.

FINDINGS
Health Insurance and Marital Status

• Among the near elderly (ages 51 to 61), widows are about twice as likely to lack health insurance coverage relative to married women (24.6 percent vs. 12.4 percent).
• Twenty percent of near elderly uninsured women are widows; 16 percent are divorcées; these groups represent 36 percent of near elderly women who are uninsured but a smaller share (26 percent) of all near elderly women.
• Near elderly women who later became widows had lower initial rates of health insurance than women whose husbands did not die; 79.3 percent of those initially married and widowed within the next ten years initially had health insurance, compared to 89.4 percent of those whose husbands remained alive for the next ten years.

Widowhood and Losing/Gaining Health Insurance

• The relationship between coverage and widowhood appears to be related to the characteristics of women likely to become widows, as opposed to a causal effect of widowhood.
• In a hazard model of insurance loss among women initially insured and married, widowhood increased the risk of becoming uninsured compared with remaining married. However, adding a small number of covariates, including age, race, ethnicity, education, and an indicator for income less than 125 percent of the applicable poverty threshold, made the effect of widowhood statistically insignificant, again suggesting that widowhood is less important than other characteristics in explaining changes in coverage.

Effect of Wife’s Employment-Based Coverage
Wives with coverage through their own employment had both a lower risk of widowhood and a lower risk of becoming uninsured if widowed.

Effect of Widowhood on Labor Force Participation
Compared to non-working women who stayed married, non-working widows are more likely to re-enter the workforce. It is not clear how demand for health insurance affects this process.

POLICY IMPLICATIONS
For most near-elderly women, widowhood does not increase the likelihood of being uninsured. The risk of becoming uninsured is greater among black women with low levels of education and very low incomes. Policies should focus on the 10-20 percent of women who lack coverage during these vulnerable years, putting them at risk for serious financial hardship or of deferring needed care until they become Medicare eligible.

CAVEATS
The study does not investigate the mechanisms by which widowed women retain coverage (such as COBRA coverage, coverage through the deceased husband’s employer, or obtaining a job with health insurance). The study does not consider how characteristics of the deceased husband may influence post-widowhood insurance status. For example, women whose husbands die because of accidents or other unanticipated causes may follow different paths than those whose husbands die after a long illness. The low divorce rate among the near elderly (n=48 for newly divorced women in the HRS sample) provides a sample too small to make precise inferences about health insurance transitions of those who divorce.

DATA SOURCE
Health and Retirement Survey original cohort, born 1931-1941; five survey waves: 1992, 1994, 1996, 1998, and 2000. Analysis focuses on women in the 1992 cohort ages 51 to 61 who were married in 1992 and who participated in all survey waves through 2000.

METHODOLOGY
Multinomial logistic regression models employed to estimate differences across insurance categories. Hazard models used to predict loss of health insurance.

CITATION
Widowhood, Divorce, and Loss of Health Insurance among Near-Elderly Women: Evidence from the Health and Retirement Study
David R. Weir and Robert J. Willis, University of Michigan

Conference paper presented at ERIU Research Conference, July 2002

ERIU Working Paper #7 (Adobe PDF)

Back to top