Midlife Work and Women’s Long-Term Health and Mortality

Jennifer Caputo, Max Planck Institute for Demographic Research
Melissa Hardy, Pennsylvania State University
Eliza Pavalko, Indiana University

While paid work is a well-established predictor of health, several gaps in our knowledge about the relationship between adult work patterns and later health and mortality remain, including whether these benefits remain stable over long periods of time and whether they are dependent on job characteristics and experiences. In this paper we draw on more than three decades of data from the National Longitudinal Survey of Mature Women to assess how labor force participation over a period of twenty years during midlife affects mental and physical health and mortality over the following fourteen to twenty-three years. We find that persistent work earlier in life continues to predict improved health and longevity many years later as women pass retirement, even after accounting for many health-linked variations in this work experiences and the presence of later life work. These findings add to knowledge about the cumulative nature through which key adult social experiences shape health as individuals enter later life.

Midlife Work and WomenÕs Long-TermHealth and Mortality

Paidwork is one of the most important social determinants of health, providingindividuals with a host of resources that protect and enhance well-being throughout the life course.  Research shows that the health benefitsof work are contingent on its stability and consistency, which life coursescholars argue reflects a pathway of cumulative advantage. However, three questionsabout how adult work histories affect individualsÕ health as their lives unfoldremain. First, even among recent studies using longitudinal data on worktrajectories, later health is typically measured at one point or over a shortfollow-up. We know little about how the health benefits of work evolve as time continues.Secondly, knowledge about how qualitative experiences with work shape itshealth effects is limited, despite studies showing that these experiencesmediate its health benefits in the short-term (e.g., House et al. 1986; Lennon1994). Finally, the extent to which the long-term health effects of work dependon the indicator examined remains unclear, since most research focuses onsingular health symptoms or mortality.

Addressingthese gaps, we draw on data from the National Longitudinal Survey of MatureWomen, a representative sample of 5,083 U.S. women who were age 28-45 when the studybegan in 1967, and re-interviewed about every two years until 2003. Ouranalysis assesses the relationship between womenÕs work over the first 20 yearsof the study and the following fourteen years of health from 1989-2003, as wellas mortality from 1989-2012. Depressivesymptoms are measured at six time points with seven items from the CES-Dscale, and Functional limitations areassessed with five items. Mortalityinformation to 2012 was linked to the survey using the National Center onHealth Statistics. Our main independent variable, Proportion of waves worked 1967-1987, measures the frequency ofwomenÕs work participation during these two decades. We also include later lifework status and several measures of the characteristics of midlife work, inaddition to sociodemographic and control variables. Weuse linear growth curve models to estimate patterns in depressivesymptoms and functional limitations from 1989-2003. To explore differences inage-specific mortality based on work histories, we employ Cox proportionalhazards regression.

Table1 presents truncated results from linear growth models regressing depressivesymptoms and functional limitations from 1989-2003 and hazards ratios fromregressions of age-dependent mortality to 2012 on work from 1967-1987. Results fordepressive symptoms show that the proportion of waves worked during these twodecades is a significant negative predictor of later depression. This is evenaccounting for most recent work status, structural aspects of work (income,hours and occupation group), and qualitative appraisals of the midlife workexperience (perceived discrimination, work attitude and commitment). Thenon-significant interaction between work and time suggests that thisrelationship stays consistent over the following years. Results for functionallimitations also show a protective effect of persistent midlife work, despitecontrolling for other longitudinal health-linked characteristics and mostproximate available work status. The interaction between work and time is againnon-significant, suggesting a stable effect. Lastly, results for age-dependentmortality to 2012 indicate that persistent midlife work predicts decreasedhazard of death, even accounting for the reduction in mortality associated withmost recent work status and the effects of other work characteristics.

Insum, our analyses show that persistent work in midlife predict improved mentaland physical health and longevity many years later as women pass retirement,even after considering the many health-linked variations in these workexperiences and the presence of later life work. We also show that thesepatterns remain stable over the next 14-23 years, supplementing studies thatdocument shorter-term benefits of work. These findings add to knowledge aboutthe cumulative processes through which adult roles and experiences shape healthas individuals enter late life.

Table 1: Coefficients for Regressions of Health in 1989-2003 and Hazards Ratios for Age-Dependent Mortality 1989-2012

Depressive Symptoms


Functional Limitations (N=2,869)












Proportion waves worked 1967-1987









Proportion waves worked 1967-1987* Time








Worked last valid interview 1989-2003









Natural log of family income 1987









Main occupation group 1987-1967 (Comparison: Blue collar)


     White collar









     Pink collar









Average hours worked 1967-1987









Proportion waves reported discrimination 1967-1987









Average positive job attitude score 1967-1987









Proportion waves reported job commitment 1967-1987









NOTE: Age, health limitations in 1967, race, education, marital status, parental status are included as controls in all models.

*=p<.05, **=p<.01, ***=p<.001

Presented in Session 1134: Life Course