Twenty-five years ago, married women's retirement decisions were strongly influenced by their husbands' health and retirement status, factors determining the value of women's nonmarket time. In contrast, their own economic opportunity set (wages, Social Security entitlements, and employer pension benefits) appeared to have little effect on their decisions to leave the labor force.' Married women currently forming expectations regarding retirement differ in important ways from this earlier generation. They have spent more time in the labor force, earned higher wages, and accumulated substantial pension rights, both private and public. They also have lower probabilities of remaining married. In 1970, 82 percent of U.S. women aged 45-54 were manried, while 5 percent were divorced. By 1992, only 73 percent were married, while 16 percent were divorced. Thus, husbands' pension and Social Security benefits are less likely to provide economic security in retirement for the current generation of preretirement married women. Their expectations regarding retirement should reflect these changing conditions. Relative to earlier cohorts, married women's retirernent plans should be more strongly influenced by considerations of their own economic returns from continued employment. While of interest for its labor-supply implications, this issue is a matter of public concern because of growing evidence that divorce has wide-ranging consequences for the economic well-being of postretirement women (William H. Crown et al., 1993). Findings from the new Health and Retirement Survey indicate that older married women's expectations of working after age 62 are strongly influenced by their expected wage, nonwage compensation such as employer-provided health and disability insurance, and pension income. Expected Social Security entitlements also appear important, although the evidence for their effect is weaker. Like the earlier generation, wives are also influenced by their husbands' plans, suggesting a tendency toward joint retirement.

If, as is usually assumed, older individuals face a continuous choice of work hours without fixed costs or take account of the actuarial adjustment of Social Security benefits postponed as a result of the earnings test, the earnings limit should not affect their labor supply before age 65. We test, and reject, these assumptions by estimating the hazard function for labor market reentry after retirement, using white men in the Retirement History Survey. We find that the earnings limit does affect reentry and that older men behave myopically, responding to current benefits rather than to Social Security wealth. Several policy implications follow.

A model of labor force participation, market wages, and labor supply of annual hours and weeks of work is estimated for married males and unmarried females in a sample from the retirement history survey, 1969-75, matched with social security earnings records. Emphasis is placed on the effects of variables associated with retirement and old age and on the estimation of net age profiles purged of biases associated with specific years and cohorts and with the other included variables. The effects of most of the usual variables appear to be consistent with the theory and with other empirical findings for both sexes. In addition, age itself has a strong net autonomous effect on each of the dependent variables due to biological, societal, or institutional factors.

Giora Hanoch, Marjorie Honig, "True" Age Profiles of Earnings: Adjusting for Censoring and for Period and Cohort Effects, The Review of Economics and Statistics, Vol. 67, No. 3 (Aug., 1985), pp. 383-394

The decline in health insurance coverage over the last two decades is a matter of national concern. The vast majority of insured individuals under age 65 obtain coverage from their employer or as dependents of a family member with group-sponsored health insurance. Recent evidence suggests that the decrease in coverage among full-time workers has resulted not so much from declining employer offers but rather from reduced take-up (Henry S. Farber and Helen Levy, 2000). The reasons for this change in employee behavior are not yet understood. Resolving this question is important for public policy because the two most likely explanations (the rise in employee insurance premiums and the increase in coverage of spouses under their own plans over this period) have different policy implications. Table 1 shows these trends by gender for full-time married workers from 1988 to 2001. The proportion of workers paying part or all of total insurance premiums increased nearly 20 percent, and the proportion with spouses covered under their own employer plans increased 30 percent among women and doubled among men. Table 1 also confirms that, while offer and eligibility rates declined somewhat for full-time workers over this period, take-up rates fell substantially among eligible workers. Concern among policymakers has focused on whether rising employee costs have forced workers to opt out of offered coverage. In this case, the number of uncovered individuals and

Homelessness in America has become a major policy concern in recent years. Following estimates by a number of researchers in the 1980's that suggested as many as a half-million homeless,1 the 1990 census places the current number at around a quarter of a million. Efforts to design policies to deal with this problem have been handicapped by a lack of systematic analysis of the causes of homelessness.2 Policymakers have had little guidance from researchers in determining the relative importance of such potential causes of homelessness as tight housing markets, slack labor markets, reductions in real publicassistance benefits, tightening of eligibility requirements for public assistance, and noninstitutionalization of the mentally ill. To assess the relative importance of these and other factors, we used estimates by the Department of Housing and Urban Development (HUD) of the homeless population in a cross-section of metropolitan areas in 1984. We do not attempt in this paper to develop a full structural model of homelessness. Rather, we present reduced-form findings as a first attempt to measure comprehensively and systematically the underlying causes of homelessness.3 We assume in this paper that homelessness results from an imbalance between the cost of available housing and a household's income. Such an imbalance may occur, for example, when housing markets are tight relative to labor markets and housing costs are therefore high relative to earnings (or to alternative resources, such as public assistance). Investigation of the causes of homelessness must go beyond housing markets alone, however, because of the special characteristics of the population at risk and the public policies that address their needs. Transfer payments and policies regarding institutionalization of the mentally ill, for example, should be important determinants of the incidence of homelessness but are not part of a standard housing model. Since homelessness represents the end of a spectrum of poor housing outcomes, we also estimate equations for two related conditions, crowded and doubled-up housing. These are often cited as causes of homelessness but are, in fact, different manifestations of the same underlying relationship between housing costs and household resources.4 * Department of Economics, Hunter College/ CUNY, 695 Park Avenue, New York, NY 10021. This research was supported by the Institute for Research on Poverty under the Small Grants Program and the University Committee on Research of the City University of New York. The authors thank Steve BartolomeiHill, Howard Chernick, Steven Craig, Martha Hill, Charles Manski, Kathryn Nelson, and Cordelia Reimers for helpful comments, and Franco Pignataro for his unusually meticulous programming assistance. 'See Filer and Honig (1990) for a detailed discussion of estimates of the size and growth of the homeless population. 2Most studies have focused on the size of the homeless population and its demographic characteristics. A few analyses have examined a limited range of potential causes of homelessness. F. Stevens Redburn and Terry F. Buss (1986) examined the roles of population growth, climate, and housing conditions; William Tucker (1987) and John M. Quigley (1990) focused on the role of rent control. Karin Ringheim (1990) analyzed in depth a small sample of metropolitan areas. 3We view this effort as preliminary to a more detailed analysis we intend to undertake when complete data from the 1990 Census, which enumerates the homeless population, become available. 4Most previous economic analyses of housing markets have not focused on the decision of whether to share housing. Demographers, on the other hand, have long studied this issue when examining family formation and living arrangements. Among economists, a notable exception is the study by Axel Borsch-Supan and John Pitkin (1988), who focused on the choice between renting, purchasing, and sharing housing. We are concerned in this paper with the other margin, where individuals and families chose between sharing housing and entering the homeless shelter system. Such