The University of Michigan
555 South Forest Street
Third Floor
Ann Arbor, MI 48104-2531

T 734-936-9842
F 734-998-6341


In a study funded by the Economic Research Initiative on the Uninsured (ERIU), Anne Beeson Royalty and Jean M. Abraham find labor market decisions of one spouse are affected by the health insurance options offered to the other spouse. Their work is innovative in how it takes into account spouses making job decisions jointly.

Descriptive statistics
Among two adult married households in the sample, 67.5% of men are offered health insurance and 45.8% of women. Among wives whose husbands were not offered, 43% were themselves offered health insurance. Among wives who do not hold health insurance in their own name, 65% had husbands eligible for coverage through their employment.

Effect of Spousal Offer on Own Offer

  • Among married couples in the sample, each 10% increase in the predicted probability of the husband being offered health insurance was associated with a 3.52 percentage point (SE: .77) decrease in the probability of the wife having an offer. In a similar model of wife's sick leave, a 10% increase in predicted probability of husband's health insurance offer is associated with a 2.02 percentage point (SE: .76) decrease of wife's paid sick leave. Royalty and Abraham argue that spousal paid sick leave confers a benefit only on the spouse who has it, and thus sick leave captures the bias that assortative mating injects into instruments often used for spouse's health insurance: spouse's job and spouse's human capital characteristics. Subtracting paid sick leave from health insurance, each 10% increase in husband's predicted offer probability lowers the wife's probability by 1.5 percentage points (bootstrapped SE: .59).
  • The analogous effect for husbands is a .10 percentage point (SE: .55) decrease in own health insurance offer for a 10% increase in wife's predicted probability of health insurance offer, and 1.76 percentage point (SE: .59) increase in the probability of paid sick leave from a 10% increase in wife's predicted health insurance probability. Again subtracting the husband's effect on sick leave from husband's effect on health insurance, the net effect of husband's offer on wife's offer is a 1.87 percentage point (SE: .52) decrease for each 10% increase in husband's offer probability.

Effect of Spousal Offer on Work

  • For wives, each 10% increase in the predicted probability of a husband working in any job with a health insurance offer reduces the probability of her working full-time in a job with a health insurance offer by .96 percentage points (SE: .50) while for husbands, the analogous effect is a 2.11 percentage point decrease (SE: .050).

Sorting by two-worker couples can blunt the effect of policy initiatives targeted at part-time workers or workers in small firms. Part time work or work for a small employer may be part of a two-worker strategy to obtain health insurance through one worker. When this is true, part time workers or workers in small firms are unlikely to respond to the policy initiative. Also, some share of those who take advantage of policies to provide insurance to part time workers or those who work for small firms will be individuals who use the policy to allow them to move from full time to part time work or from working for a large employer to working for a small employer.

Increases in co-premiums charged for family coverage may cause declines in coverage in the short run because many families have sorted into jobs such that only one spouse has coverage. Labor supply and choice of employment opportunities will be affected by changes in the availability and price of family coverage.

Instrumental variables: Use of sick leave as an instrument in order to identify the behavioral effect of spouse's insurance on own insurance. While a health insurance offer confers a benefit to both spouses, paid sick leave can only be used by the spouse who has the benefit. Royalty and Abraham argue that, in a model of own health insurance, the bias on spouse's characteristics as an instrument for spouse's health insurance offer can be captured by the coefficient on spouse's health insurance status in a model estimating own paid sick leave. Assortative mating would produce an upward bias on the instrument and shared household income a downward bias.

Models estimated as linear probability models, an indicators for non-white, total number of children age 18 or younger in the household, number of serious medical conditions reported by all household members, and income from dividends and interest, linear and quadratic terms for age and education, regional dummies, year indicators, and unemployment rate for the county of residence.

The results rest on acceptance of the identifying assumption: that sick leave captures the correlation between spouses that otherwise would bias spouse's characteristics as instruments. Like any argument from asymptotic properties of an estimator, whether it holds in small samples, such as considered in the paper, is unknown.

Household Component (HC) of the Medical Expenditure Panel Survey (MEPS) for 1996, 1997, and 1998. Sample is married households in which both partners are between 19 and 64, non-disabled, and not full-time students; sample is limited to married households in which at least one partner is employed outside the home. The sample contains 6,782 one- and two-earner households. The two-earner sub-sample is 4,491 households.

Health Insurance and Labor Market Outcomes: Joint Decision-Making within Households
Anne Beeson Royalty, Indiana University Purdue University Indianapolis and Jean M. Abraham, University of Minnesota

Conference paper presented at ERIU Research Conference, July 2004

The final paper is forthcoming in the Journal of Public Economics.

ERIU Working Paper #28 (Adobe PDF)

Back to top

Funded by The Robert Wood Johnson Foundation, ERIU is a five-year program shedding new light on the causes and consequences of lack of coverage, and the crucial role that health insurance plays in shaping the U.S. labor market. The Foundation does not endorse the findings of this or other independent research projects.