Examining the mortality of young baby boomers by age 58: evidence from the NLSY79

This article examines mortality in the National Longitudinal Survey of Youth 1979 (NLSY79), a sample of individuals born in the years 1957–64. The first interviews for this cohort were conducted in 1979. We restrict our analysis to mortality up to age 58, as all sample members have turned 58 by 2022, the most recent interview available to the public. By age 58, almost 10 percent of the sample is deceased: around 11 percent of men and around 8 percent of women. We examine the age distribution of mortality for these young baby boomers and analyze how demographics, early test scores, and several characteristics collected at age 25 (education, employment, and health) relate to mortality later in life. Results from Cox proportional hazard models indicate that being female, obtaining higher levels of education, earning higher test scores, and working a higher number of weeks in the year that the individual turned age 25 are all related to a lower likelihood of being deceased by age 58. In contrast, health characteristics at age 25—smoking daily, obesity, and health limitations that affect work—are related to a much higher likelihood of mortality by age 58.

Research on mortality

Research on the impact of health behaviors and socioeconomic factors on early mortality provides useful insights for public health and workforce programs, policymakers, and the public. These studies shed light on a variety of risk factors related to mortality, helping to inform intervention programs aimed at improving population health and well-being. To this end, numerous studies have investigated differences in mortality across socioeconomic status, education, and health. A substantial body of research finds that factors such as income and education are positively related to longevity and that these differences have widened over time.¹ Joblessness and poor working conditions have also been linked to health disparities and higher mortality.²

The National Longitudinal Survey of Youth 1979 (NLSY79) has historically been a key source of information for examining health disparities. For example, NLSY79 data have been used to study the intersection of health and mortality and education, poverty, and incarceration.³ As the NLSY79 sample members continue to age, mortality differences within the cohort born in 1957–64 are becoming starker. The NLSY79 provides a unique set of information about respondents throughout their lives, such as cognitive test scores, health behaviors, and detailed education and work histories. This information can be used to research the underlying drivers of mortality differences as this cohort ages.

This article examines the measurement and patterns of mortality in the NLSY79 by age 58. We first discuss NLSY79 mortality measures and compare patterns of mortality with Social Security Administration (SSA) life tables. We then summarize patterns of mortality for the cohort born in 1957–64 across a variety of characteristics and, subsequently, estimate Cox proportional hazard models to analyze the relationship between mortality by age 58 and demographics, early test scores, and age 25 characteristics. We conclude with potential directions for researchers interested in using these data, highlighting that the detailed life-history data in the NLSY79 allow for explorations of how a wide range of life events and trajectories relate to mortality.

NLSY79 mortality data

The NLSY79 began in 1979 and is a nationally representative sample of 6,403 men and 6,283 women born between 1957 and 1964 and who lived in the United States during the initial survey. Respondents were interviewed annually through 1994 and biennially afterward. By the 2022 survey, NLSY79 respondents were aged 58 to 65. The survey originally included oversamples of Black and Hispanic individuals; military personnel (dropped after 1984); and low-income non-Black, non-Hispanic individuals (dropped after 1990). After excluding the two dropped oversamples, the sample size is 9,964.

In this article, we define mortality based on the “reason for non-interview” (RNI) variable available for each round. As part of the regular fielding of the NLSY79 survey, NORC at the University of Chicago, a contractor to BLS, verifies whether a respondent is characterized as deceased.⁴ This process includes regular locating searches for all sample members, as well as specific verification of any reported death against public records, including the Social Security Death Index, credit bureau data, and reported obituaries or other announcements.

In this article, we define “age deceased” as the age the respondent turned in the interview year that he or she is first reported as deceased. Of the 9,964 respondents in our sample, 1,040 were deceased by age 58 (634 men and 406 women), or about 10 percent of the sample. By the 2022 interview, when respondents were ages 58–65, an additional 308 respondents were reported as deceased—1,348 in total. Unless otherwise specified, all analyses in this article weight data using round 1 weights to make the sample representative of the population from which the NLSY79 is drawn.

In addition to this RNI mortality information, in 2019, NLSY79 data were submitted to the National Center for Health Statistics’ National Death Index (NDI) to attempt to link deceased respondents to existing death certificates.⁵ After adjudication, confirmation (as of 2018) was determined for 858 respondents. Chart 1 compares the age of death based on RNI in the NLSY79 for matched respondents to the NDI-confirmed deaths, using unweighted data.

About 48 percent of individuals had the same age of death for both measures, about 43 percent had an RNI age of death that was a year different from the NDI-confirmed death, and almost 7 percent had an RNI age of death that was 2 years different.⁶ Of the 490 respondents who were deceased according to the NLSY79’s RNI, but not matched as deceased in the NDI, the majority (about 62 percent) were first shown as deceased in the 2020 NLSY79 interview (143 respondents) or the 2022 interview (163)—that is, after the matching to NDI had occurred. Chart 2 shows the weighted percentage deceased, cumulative by age, for the portion of the sample returned as deceased in the NDI. The chart shows proportion deceased using both the NDI age measure and the RNI age measure. Overall, the age-deceased data match well, with the RNI and the NDI measures nearly overlapping each other.

From the ages of the midtwenties to later forties, both the NDI- and RNI-weighted cumulative percent deceased for men were well above the NDI- and RNI-weighted cumulative percent deceased for women. From about the midthirties to midforties, the gap between the cumulative percent deceased of men and that of women rose above 10 percentage points.⁷

We also compare NLSY79 survival estimates with the SSA cohort-based actuarial life tables, which provide life expectancy estimates conditional on surviving to a given age by sex. We calculate survival estimates from SSA estimates averaged across NLSY79 birth cohorts and compare them with NLSY79-weighted Kaplan-Meier survival estimates.⁸ Table 1 shows SSA cohort life tables and weighted NLSY79 estimates of the cumulative percentage of surviving individuals.

Based on the data in table 1, the SSA and NLSY79 estimates were nearly identical from ages 25 to 55. For example, survival rates at age 40 for both data sources were about 96 percent for men and 98 percent for women. At age 55, the survival rate estimates in both data sources were about 90 percent for men and 94 percent for women. The similarity of these NLSY79 and SSA estimates is encouraging and reinforces the representativeness of the sample of the NLSY79, in this respect.

Table 2 uses weighted NLSY79 data to show deaths over the lifespan up to age 58 by sex, race, and ethnicity. The mortality rate for the cohort born in 1957–64 steadily increases up to age 58. By age 58, almost 10 percent are deceased.

The mortality rate for men is higher than that for women over the age ranges: by age 58, over 11 percent of men are deceased, compared with almost 8 percent of women. Black men have higher mortality rates: by age 58, over 16 percent of Black men are deceased, compared with around 10 percent of non-Black, non-Hispanic men and about 12 percent of Hispanic men. From ages 23 to 29, the mortality rate for Black men is almost three times higher than that for non-Black, non-Hispanic men and is over twice as high from ages 30 to 39. The mortality rate for Black women becomes noticeably higher than that of non-Black, non-Hispanic and Hispanic women from ages 40 to 49 and 50 to 58. By age 58, the mortality rate of Black women is almost 11 percent, compared with about 7 percent for non-Black, non-Hispanic women and around 6 percent for Hispanic women. Black women’s mortality rate at age 58 is about 5 percentage points lower than that of Black men at that age.

In the empirical analysis below, we use NLSY79 data to estimate hazard models to examine the relationship between the risk of mortality by age 58 and individuals’ demographics, their early math and verbal aptitude scores, and their age 25 characteristics. In these analyses, we control for early math and verbal aptitude with percentile scores from the Armed Forces Qualification Test (AFQT), which was given to NLSY79 respondents in 1980.⁹ Variables related to educational attainment, employment, and health were measured at age 25. Because the analysis includes controls for characteristics at age 25, we exclude the 69 individuals deceased before 25, reducing the sample size to 9,895. The weighted percent deceased is very similar in the sample of 9,964 respondents and the reduced sample of 9,895 (which excludes those deceased prior to age 25). In the full weighted sample, 11.3 percent of men and 7.7 percent of women are deceased by age 58, compared with 10.5 percent of men and 7.4 percent of women in the reduced sample. For the remainder of the analysis in this article, we use the reduced weighted sample of 9,895 respondents.

Mortality by demographics, AFQT percentile score, and age 25 characteristics

Table 3 shows summary statistics for the percent deceased by age 58 across demographics, AFQT percentile score, and age 25 characteristics. Those with the lowest quartile of AFQT percentile scores have a much higher likelihood of being deceased by age 58 than those with higher quartiles. For example, almost 15 percent of those in the lowest quartile of AFQT percentile scores are deceased by age 58, compared with about 6 percent of those with the highest quartile. The percentage-point differential between the likelihood of being deceased in the highest and lowest AFQT score quartile is more pronounced for men than for women (about 11 percentage points and 7 percentage points, respectively).

The percent deceased by age 58 declines as educational attainment increases: 17.3 percent of those with less than a high school diploma, 9.3 percent for those with a high school diploma, 7.4 percent for those with some college, and 4.6 percent for those with a bachelor’s degree or higher. The percent deceased also differs across health characteristics at 25. Those with health limits that affect the kind or amount of work they can do are about twice as likely to be deceased at age 58 than those with no health limits (about 17 percent compared with almost 9 percent). The differential is much larger for men (almost 24 percent compared with about 10 percent) than for women (almost 14 percent compared with 7 percent). Those who were obese at age 25 and those who ever smoked daily by age 25 generally have a much higher likelihood of being deceased by age 58 than their counterparts without these characteristics.

Charts 3 and 4 show the cumulative percent deceased over time by AFQT percentile-score quartile and education level at age 25. Chart 3 shows a growing differential based on AFQT score as the cohort born in 1957–64 ages.

Most notably, respondents with scores in the lowest quartile of AFQT have markedly higher mortality rates than those in the higher three quartiles, and this gap continues to expand as the cohort has aged. Through about the midforties, the cumulative percent deceased in the higher three quartiles was under 4 percent. After the midforties, the cumulative percent deceased in the second lowest quartile of AFQT scores rose above the two highest quartiles of scores.

Chart 4 shows similar results for education at age 25, with a very pronounced differential between cumulative percent deceased for respondents with less than a high school education and those with at least a high school level education. This begins to noticeably increase in the midthirties, then again in the later thirties, and again in the early fifties to midfifties.

Results: factors related to the risk of mortality by age 58

We estimate weighted Cox proportional hazard models to examine the relationship between demographics, early test scores, and age 25 characteristics and the risk of mortality by age 58. In tables 4 (all), 5 (men), and 6 (women), we show estimates of hazard ratios, standard errors, and significance levels (p-values) from five specifications that each build on the previous one by adding additional controls. Specification 1 includes basic demographic controls; specification 2 adds AFQT test score; specification 3 adds education at age 25; specification 4 adds employment and other characteristics at age 25; and, finally, specification 5 adds controls for health at age 25. A hazard ratio greater than 1 indicates the variable is associated with a relatively higher risk of death by age 58, and a hazard ratio less than 1 indicates the variable is associated with a relatively lower risk of death by age 58. Our results are descriptive and should not be interpreted as causal.

In table 4, specification 1 shows that being female is associated with a lower risk of mortality by age 58 (about 31 percent) relative to being male. Being Black is associated with a much higher risk of mortality by age 58 relative to being non-Black, non-Hispanic. Relative to the highest quartile of AFQT scores, the lowest two quartiles of AFQT scores are associated with a higher risk of death by age 58 (around 180 percent for the lowest quartile and almost 72 percent for the second lowest quartile). With the addition of educational attainment at age 25 (specification 3), the hazard ratio for the second lowest quartile of AFQT scores is no longer statistically significant, and the magnitude of the lowest quartile decreases to around a 71-percent relative risk. Relative to the lowest level of education (less than a high school diploma), all higher levels of education are associated with a statistically significant lower risk of mortality by age 58. For example, having a bachelor’s degree or higher is associated with an approximately 63-percent lower risk of mortality by age 58, relative to having less than a high school education.¹⁰

Specification 4 of table 4 adds controls for employment, marital status, the presence of the respondent’s own children (of any age) in their household, and whether the respondent resided in an urban location at age 25. Both education and the lowest quartile of AFQT scores remain statistically significant predictors of mortality, and their magnitudes are similar to those in specification 3. Working fewer than 45 weeks in the year the individual turned age 25 is associated with an increased risk of mortality by age 58. Being married at age 25 is associated with an almost 30-percent lower risk of mortality by age 58, relative to not being married at age 25.

Specification 5 adds controls for health measures at age 25. The coefficient estimates and statistical significance of education, employment, and marital status measures at age 25 from specification 4 remain similar in specification 5. In addition, being female continues to be associated with a lower risk of mortality by age 58, and having the lowest quartile in AFQT scores continues to be associated with a higher risk. Unsurprisingly, the results in this specification show that health measures at age 25 are associated with the risk of mortality by age 58. For example, reporting health limits on the kind or amount of work an individual can do at age 25 is associated with an almost 100-percent higher risk of mortality by age 58, relative to reporting no health limitation, and being obese at 25 is associated with an almost 74-percent higher risk of mortality by age 58, relative to being a healthy weight. Having ever smoked daily by age 25 is associated with an almost 69-percent higher risk of mortality by age 58, relative to not having ever smoked daily. Tables 5 and 6 show results from Cox proportional hazard models estimated separately by sex.

As with the full sample, once additional controls are included, race is no longer statistically significant for men or women. However, in specifications with more covariates, being Hispanic is associated with a lower risk of mortality by age 58 for women, but not men, relative to being non-Black, non-Hispanic. Results for both men and women show a higher risk of mortality for those with the lowest quartile AFQT scores, relative to the highest.

Once health measures are added to specification 5, education measures are no longer statistically significantly associated with the risk of mortality by age 58 for women but remain so for men. For both men and women, working less than 45 weeks at age 25 is associated with a higher risk of mortality at age 58 than working 45 weeks or more. Living in an urban location at age 25 is related to a lower risk of mortality at age 58 for women, but not for men, and being married at age 25 is associated with a lower risk of mortality for both men and women. Finally, having health limitations, being obese, and ever smoking daily by age 25 are all associated with a higher risk of mortality by age 58 for both men and women.

Directions for future research

This article provides a high-level overview of mortality patterns using the NLSY79 data, highlighting its potential for understanding long-term health outcomes. We find that 11.3 percent of men and 7.7 percent of women were deceased by age 58 in the NLSY79 cohort of young baby boomers. NLSY79 and SSA survival rate estimates by sex are remarkably similar, lending credence to the analysis of mortality with NLSY79 data. Cox proportional hazard models with controls for demographics and age 25 characteristics show that being male, having lower cognitive test scores, having lower levels of education, and working fewer than 45 weeks at age 25 are related to an increased risk of mortality by age 58. Health at age 25 is highly related to mortality by age 58: obesity, ever having smoked daily, and having a health limit that affects the kind or amount of work you can do are associated with an increased risk of mortality by age 58.

While this article focuses on high-level patterns, the richness of the NLSY79 provides the ability for a wide array of future research opportunities. For example, researchers can explore impacts of life events such as health shocks, divorce, and job loss on mortality. Similarly, future research could explore how occupational histories, hours fluctuations, and other work characteristics shape health over time. Beyond employment and life events, the NLSY79 enables investigation into other critical domains that influence mortality. Educational experiences and skill accumulation; patterns of wealth and debt; and broader social determinants, such as housing and neighborhood environments, are all captured in the data. Importantly, researchers are not only able to explore how these factors individually predict mortality but can also explore the complex interactions between these factors using the cross-domain nature of the NLSY79 data.

Because the NLSY79 tracks individuals across decades and domains, researchers can study both the timing of events (for example, does job loss have different effects at various life stages?) as well as how trajectories of life circumstances predict mortality (for example, how do occupational pathways mediate the relationship between education and mortality?). This combination of breadth and depth makes the NLSY79 a valuable resource for investigating the complex dynamics underlying the determinants of health across the life course.

References

Berkman, Lisa F., Ichiro Kawachi, and Maria M. Glymour. Social Epidemiology. 2nd ed. Oxford, England: Oxford University Press, 2014.

Bernstein, Shayna F., David Rehkopf, Shripad Tuljapurkar, and Carol C. Horvitz. “Poverty Dynamics, Poverty Thresholds and Mortality: An Age-Stage Markovian Model.” PLOS ONE 13, no. 5 (2018). https://doi.org/10.1371/journal.pone.0195734.

Bosworth, B., Gary Burtless, and Kan Zhang. Later Retirement, Inequality in Old Age, and the Growing Gap in Longevity Between Rich and Poor. Washington, DC: Brookings Institution, 2016.

Bovell-Ammon, Benjamin J., Ziming Xuan, Michael K. Paasche-Orlow, and Marc R. LaRochelle. “Association of Incarceration with Mortality by Race from a National Longitudinal Cohort Study.” JAMA Network Open 4, no. 12 (December 1, 2021). https://doi.org/10.1001/jamanetworkopen.2021.33083.

Braveman, Paula A., Catherine Cubbin, Susan Egerter, David R. Williams, and Elsie Pamuk. “Socioeconomic Disparities in Health in the United States: What the Patterns Tell Us.” American Journal of Public Health 100, supplement 1 (2010), pp. S186–S196. https://doi.org/10.2105/ajph.2009.166082.

Bundy, Joshua D., Katherine T. Mills, Hua He, Thomas A. LaVeist, Keith C. Ferdinand, Jing Chen, and Jiang He. “Social Determinants of Health and Premature Death Among Adults in the USA From 1999 to 2018: A National Cohort Study.” Lancet Public Health 8, no. 6 (2023), pp. e422–e431. https://doi.org/10.1016/s2468-2667(23)00081-6.

Burgard, Sarah A., and Katherine Y. Lin. “Bad Jobs, Bad Health? How Work and Working Conditions Contribute to Health Disparities.” American Behavioral Scientist 57, no. 8 (August 2013). https://doi.org/10.1177/0002764213487347.

Case A., and Angus Deaton. "Mortality and Morbidity in the 21^st Century." Brookings Papers on Economic Activity 2017, no. 1 (Spring 2017), pp. 397–476. https://doi.org/10.1353/eca.2017.0005.

Chetty, Raj, Michael Stepner, Sarah Abraham, Shelby Lin, Benjamin Scuderi, Nicholas Turner, Augustin Bergeron, and David Cutler. “The Association Between Income and Life Expectancy in the United States, 2001–2014.” Journal of the American Medical Association, 315, no. 16 (Apr 26, 2016), pp. 1750–66. https://doi.org/10.1001/jama.2016.4226.

Conti, G., James Heckman, and Sergio Urzua. “The Education-Health Gradient.” American Economic Review 100, no. 2, (2010), pp. 234–238. https://doi.org/10.1257/aer.100.2.234.

Cristia, Julian P. The Empirical Relationship Between Lifetime Earnings and Mortality. (U.S. Congressional Budget Office, 2007). https://www.cbo.gov/sites/default/files/110th-congress-2007-2008/workingpaper/2007-11_0.pdf.

Cutler D., Angus Deaton, and Adriana Lleras-Muney. “The Determinants of Mortality.” Journal of Economic Perspectives 20, no. 3 (2006), pp. 97–120.

Daza S., Alberto Palloni, and Jerrett Jones. “The Consequences of Incarceration for Mortality in the United States.” Demography 57, no. 2 (April 2020), pp. 577–598. https://doi.org/10.1007/s13524-020-00869-5.

Duggan, James E., Robert Gillingham, and John S. Greenlees. “Mortality and Lifetime Income: Evidence from US Social Security Records.” IMF Staff Papers 55, no. 4 (2008), pp. 566–594. https://www.imf.org/en/publications/wp/issues/2016/12/31/mortality-and-lifetime-income-evidence-from-u-s-20205.

Jokela, M., Marko Elovainio, Archana Singh-Manoux, and Mika Kivimäki. “IQ, Socioeconomic Status, and Early Death: The US National Longitudinal Survey of Youth” Psychosomatic Medicine 71, no. 3, pp. 322–328. https://doi.org/10.1097/PSY.0b013e31819b69f6.

Kitagawa, Evelyn M., and Philip M. Hauser. Differential Mortality in the United States: A Study in Socioeconomic Epidemiology. Cambridge, MA: Harvard University Press, 1973.

Mackenbach, Johan P., Irina Stirbu, Albert-Jan R. Roskam, Maartje M. Schaap, Gwenn Menvielle, Mall Leinsalu, Anton E. Kunst, and European Union Working Group on Socioeconomic Inequalities in Health. “Socioeconomic Inequalities in Health in 22 European countries.” New England Journal of Medicine 358, no. 23 (Jun 5, 2008), pp. 2468–81. https://doi.org/10.1056/nejmsa0707519.

Marmot M., “Social determinants of health inequalities.” Lancet 365, no. 9464 (2005), 1099–1104. https://doi.org/10.1016/s0140-6736(05)71146-6.

Meara, Ellen R., Seth Richards, and David M. Cutler. “The Gap Gets Bigger: Changes in Mortality and Life Expectancy, By Education, 1981–2000". Health Affairs 27, no. 2 (2008), pp. 350–360. https://doi.org/10.1377/hlthaff.27.2.350.

National Academies of Science, Engineering, and Medicine. The Growing Gap in Life Expectancy by Income: Implications for Federal Programs and Policy Responses, 2015. https://doi.org/10.17226/19015.

Olshansky, Jay S., Toni Antonucci, Lisa Berkman, Robert H. Binstock, Axel Boersch-Supan, John T. Cacioppo, Bruce A. Carnes, Laura L. Carstensen, Linda P. Fried, Dana P. Goldman, James Jackson, Martin Kohli, John Rother, Yuhui Zheng, and John Rowe. “Differences in Life Expectancy Due to Race and Educational Differences are Widening, And Many May Not Catch Up.” Health Affairs (Project Hope) 31, no. 8 (2012), pp. 1803–1813. https://doi.org/10.1377/hlthaff.2011.0746.

Swift, Samuel L., Zihan Chen, Calvin Colvin, Katrina Kezios, Sebastian Calonico, and Adina Zeki Al Hazzouri. “Unsecured Debt in Early Adulthood and Premature Mortality in Adults in the USA: A Longitudinal Analysis of Prospective National Cohort Data." Lancet 10, no. 11 (2025), pp. 979–987. https://doi.org/10.1016/S2468-2667(25)00226-9.

Tapia Granados, José A., James S. House, Edward L. Ionides, Sarah Burgard, and Robert S. Schoeni. “Individual Joblessness, Contextual Unemployment, and Mortality Risk,” American Journal of Epidemiology 180, no. 3 (August 1, 2014) pp. 280–287. https://doi.org/10.1093/aje/kwu128.

Waldron, H., “Mortality Differentials by Lifetime Earnings Decile: Implications for Evaluations of Proposed Social Security Law Changes” Social Security Bulletin. 73, no. 1 (2013), pp. 1–37.

Woolf, Steven H., and Heidi Schoomaker. "Life expectancy and mortality rates in the United States, 1959-2017." Journal of the American Medical Association, 322, no. 20 (2019), 1996–2016. https://doi.org/10.1001/jama.2019.16932.