Lifetime years married held steady for men with a BA degree since 1960 but dropped to lowest level since 1880 for men without a BA

Significance

The number of years Americans spend married and in other marital statuses across their lives has major consequences for their lived experiences. In the 1980s and 1990s, there was much debate among scholars and the public about declining marriage rates, but research showed that, in terms of expected lifetime years married, surprisingly little change had occurred between the Baby Boom era and 1980. We take a long-run view, estimating expected lifetime years married and in other marital statuses from 1880 to 2019. We find striking declines in expected lifetime years married since 1960 for men and especially for men without a BA. By contrast, expected lifetime years married for men with a BA has remained high and relatively stable.

Abstract

Trends in life expectancy and marriage patterns work together to determine expected lifetime years married. In 1880, adult life expectancy was short and marriages were more likely to end by death than divorce. Since then, although there have been substantial life expectancy gains in adulthood, marriage has been increasingly delayed or forgone and cohabitation and divorce are far more prevalent. Whether adults today can expect to spend more or fewer years married than in the past depends on the relative magnitude of changes in mortality and marriage. We estimate trends in men’s expected lifetime years married (and in other marital statuses) from 1880 to 2019 and by bachelor’s degree (BA) status from 1960 to 2019. Our results show a rise in men’s expected lifetime years married between 1880 and the Baby Boom era and a subsequent fall. There are large and growing differences by BA status. Men with a BA have had high and relatively stable expected lifetime years married since 1960. For men without a BA, expected lifetime years married has plummeted to lows not seen among men since 1880. Cohabitation accounts for a substantial fraction, although not all, of these declines. Our results demonstrate how increasing inequality in both life expectancy and marriage patterns combine to amplify educational differences in lifetime experiences of coresidential partnerships.

Increased life expectancy and changes in marriage have had major consequences for American life. In 1880, life expectancy was short and marriages were more likely to end by death than divorce. Since then, adult life expectancy has increased substantially, increasing the potential number of years available to spend married. But, reductions in marriage—from high rates of separation and divorce, record-high ages at first marriage, and increased rates of nonmarital cohabitation—work in the other direction, reducing the proportion of adults’ lives spent married. Whether adults spend more or fewer years married today than in the past depends on the relative magnitude of these two forces. If gains from life expectancy have outpaced reductions in marriage, then lifetime years married will have increased. If gains from life expectancy have lagged behind reductions in marriage, then lifetime years married will have declined.

Past research has shown surprising results about the combined effects of changes in mortality and marriage patterns between 1800 and 1980 (1). Although there has been much concern at various times from the public and in academia about reductions in marriage, the expected number of years married in adulthood was greater in 1980 than it was in earlier eras with the exception of the heyday of marriage during the Baby Boom era. This finding represents an important counterpoint to claims about marriage decline. As Watkins, Menken, and Bongaarts (1:353) argue, “If number of years spent in these family statuses is taken as a rough measure of investment in the family, clearly this investment has increased, not decreased.”

Watkins et al.’s (1) and others’ research has demonstrated that the postwar Baby Boom era was an exceptional period relative to what came before and what came after (e.g., refs. 2 and 3), but are we now living through an equally exceptional period? Since 1980, there have been further gains to adult life expectancy, age at first marriage for men and women has reached record highs, and cohabitation and divorce remain common. Thus, it is an open question whether lifetime years married is at a record high, low, or somewhere in between. A long-run view of family and demographic change is necessary to put what is currently happening in family life into perspective. In this article, we estimate trends in expected lifetime years married and in other marital statuses from 1880 to 2019 in adulthood, defined as age 15+ given the young age at marriage in some of the historical periods our time series covers. Watkins et al.’s last data point was 1980—a very different context than the present day.

Another unanswered question is how expected lifetime years married varies by socioeconomic status. The extreme level of recent U.S. inequality has been compared to the Gilded Age of the late 19th century (4, 5). Not only is income inequality very high but marked inequality in life expectancy and diverging family patterns characterize the last several decades (6, 7). The dividing line appears to be a bachelor’s degree (BA), with those with a BA or more having better health and longer lives, higher marriage rates, and more stable marriages than those without a BA (7, 8). Thus, in addition to long-run trends since 1880, we estimate shorter-run trends in expected lifetime years married for those with and without a BA from 1960, the earliest year for which we have BA-status estimates and a time in which obtaining a college degree was relatively rare, through 2019.

In this article, we focus on men because of the health crisis and increased mortality especially among men at the bottom of the class structure and concomitant reductions in marriage in recent decades.^* “Deaths of despair,” often classified as suicide, drug overdose, and alcohol-related deaths, have substantially increased among less educated Americans, and rates of death attributable to these causes among men are roughly two to three times those for women (7, 9). In addition to deaths of despair, trends in other causes of death, such as heart disease and cancer, have also contributed to the widening education gap in mortality (10). The increased mortality of those without a BA has been described as part of a cluster of trends associated with economic dislocation and stress including declining wages and fractured family and community ties (e.g., refs. 7 and 11). Estimates of the widening gulf in adult life expectancy by education have not been combined with changes in marriage to estimate divergence in expected lifetime years married. Thus, a key contribution of this article is that it brings together research on education differences in mortality and marriage to estimate lifetime experiences in various marital statuses.

Lifetime years married and in other marital statuses may have substantial consequences for individuals’ lives. For example, marriage is associated with a host of positive outcomes, but many are likely not causal, and research is ongoing. We primarily view trends in lifetime years married as an indicator or outcome of larger structural socioeconomic shifts. It is well known that the likelihood of marriage and other relationship transitions depends heavily on individuals’ economic position (12). But in addition, recent research using rigorous methodology to identify causal effects shows that marriage may have positive effects in some areas. Particularly among men, marriage has been found to reduce drug and alcohol consumption, protect against criminal behavior, and improve some measures of physical and emotional health (13–16). There is less research on the effects of cohabitation; some suggests little difference between cohabitation and marriage across a range of outcomes (17). Other research suggests that cohabitation too may have positive, albeit weaker, effects on health behaviors (13). A key difference between cohabitation and marriage is that cohabiting unions tend to be shorter and are less institutionalized, which may affect the likelihood of long-term joint investments (18). Whatever the effects of marriage and other marital statuses, if they compound over time, longer exposure may be more consequential than shorter exposure (19, 20). Given the connection to duration of exposure, our focus is on lifetime years married and in other marital statuses. We present trends in the proportion of adults’ lives for comparison in SI Appendix, section B.^† Regardless of the causal effects of marital statuses, the lived experiences of people in different marital statuses are distinct in many ways. Patterns of time use, spending, homeownership, childbearing, social networks, leisure, and housework, among other differences vary by marital status, in ways not always in favor of marriage (17, 21–26).

Materials and Methods

We use two types of data to estimate men’s expected adult lifetime years married and in other marital statuses: 1) age- and year-specific mortality rates from period life tables for the total population of U.S. men, which are used by demographers to calculate life expectancy, and 2) age- and year-specific estimates of the distribution of adult men’s marital status. Our estimates of age- and year-specific full-population male mortality from 1880 to 2019 are from published life tables from a variety of sources [1880, 1900: (ref. 27: Series I pp. 157, Series III pp. 165); 1910, 1920, 1930: (ref. 28: table 6); 1940 to 2019: ref. 29]. We use unabridged life tables (those that contain age-specific mortality for single years of age) for all years except 1880 and 1900, for which only life tables for age groups are available.

Our estimates of men’s age- and year-specific marital status distributions are from 1880 to 2010 U.S. decennial censuses and the 2001 to 2019 American Community Survey (30). We begin our time series in 1880 because this is the first year that the U.S. census included information about marital status. The year 1890 is not included in our analysis because most of the census data in that year were destroyed by fire. We differentiate five, mutually exclusive marital statuses: never married, cohabiting (includes never and previously married cohabitors), married (includes first and later marriages), divorced/separated, and widowed. Same-sex married and cohabiting couples are included in these estimates. Cohabitors are not separately identified until 1990 because direct measures were not collected until then in census data. Marital status distributions by age and year are weighted using census/ACS data-provided person weights so that estimates are representative of the population.

Our measure of expected years lived married and in other marital statuses is conceptually similar to demographers’ typical measure of life expectancy. Period estimates of life expectancy are calculated using age-specific mortality rates for each year. Period life expectancy is equal to cohort life expectancy at birth if people born in that year experience the age-specific mortality conditions that exist in that year throughout their life. This is unrealistic because age-specific mortality rates change across birth cohorts’ lifetimes. However, long-run trends in period measures of life expectancy track those estimated from real cohorts closely, albeit with a lag (31). In periods of rapidly changing mortality, differences in life expectancy can be interpreted as changes in average years of life lost or gained (32). Regardless of whether the life expectancy of future cohorts maps onto period estimates, period life expectancy is an extremely useful summary measure of prevailing mortality conditions and trends (32, 33). The benefits and limitations of our period estimates of expected lifetime years married are the same as those for life expectancy.

We use the Sullivan method to estimate expected adult years lived married and in other marital statuses (34). The Sullivan method has been commonly used to estimate healthy life expectancy and, increasingly, other kinds of life expectancy (35–38). Unlike multistate life table methods (39), the Sullivan method does not require age- and year-specific data on entry and exit rates from the marital states of interest, a difficult requirement given the limitations of both historical and current data. In a comprehensive statistical and conceptual review of the Sullivan method, Imai and Soneji (35) found that the Sullivan method performs well relative to multistate life table methods, requires only the stationarity assumptions inherent in analyses of life expectancy using period life table methods, and allows for unbiased and consistent estimation given those assumptions. Indeed, our estimates using the Sullivan method replicate others’ using multistate life table methods well (1, 40).

The Sullivan method allocates estimated years lived from period life tables to marital states based on the proportion of the population in a given marital status by age and year. For instance, if a life table for year y indicates that there were 100,000 person-years lived among men from exact age 25 to less than exact age 26 ( ${}_1{L}_{25}^y$ ) and census/ACS data indicate that 10% of men were currently married in this age interval in year y, then we estimate that 10,000 person-years were spent married in year y for the 1-y age interval beginning at exact age 25. Expected adult lifetime years married in year y is the sum of person-years married at each age, divided by the total number of men who survived to adulthood. For this analysis, we define adulthood as beginning at age 15 as this is roughly the youngest age men married across this historical period and thus, we estimate expected adult years married and in other marital statuses at age 15. The calculation of average expected years in these five marital statuses includes all men, even those who never experienced the marital status.

We also present trends by BA/non-BA status beginning in 1960. Mortality data by BA status are not available prior to this time. In 1960, mortality ratios by sex, age group, and BA status are from Kitagawa and Hauser (ref. 41: table 2.1). For 1979-1985, they are from Rogot et al. (42), and for simplicity, we assume that these data represent conditions in 1980. From 1990 to 2019, we construct life tables by sex, single years of age, and BA status from death certificate data available through the Multiple Cause-of-Death Mortality database from the National Center for Health Statistics (43) and use denominator estimates from the 1991 to 1999 Current Population Survey and the 1990, 2000 to 2019 U.S. Census/American Community Survey (30) following the method outlined by Case and Deaton (33). Only a small portion of men received a college degree in 1960 and before (8% of men aged 15+ in 1960) and thus the results for the male population before 1960 are likely very similar to those for men without a BA. We apply sex-, age-, year-, and BA-status-specific marital status distributions to the life table estimates of person-years lived by sex, age, year, and BA status using the Sullivan method.

Scholars often present estimates by BA status for those aged 25 and older because education is still in progress at younger ages. Following this convention is not ideal for our analysis, however, because of the importance of years spent married prior to age 25. In the 1950s and 1960s, for example, it was not uncommon for men to marry in their late teen years. Beginning the analysis at age 25 would thus underestimate lifetime years married in this era. To address this, we use BA-specific mortality and marital status distributions beginning at age 21 and assume that all men share the same age-specific mortality rates and marital status distributions from age 15 to 20, regardless of their current or ultimate education. Our conclusions are robust to different assumptions about mortality and marital status by BA status in this youngest age range (SI Appendix, section C). In addition, because of sparse data on marriage distributions at the oldest ages (above age 89), we linearly impute marital status distribution missing values between ages and, when marital status distributions are missing above a given age, we assume that missing values are equal to the second to last nonmissing value. We show that our results are not sensitive to this imputation procedure in SI Appendix, section D.

To estimate the extent to which changes in mortality can account for trends in expected lifetime years married, we present the results of counterfactual trends holding age- and BA-status-specific mortality rates constant at their 1960 levels but allowing marriage patterns to vary as observed. To do this, we use the Sullivan method as described above, but rather than using the observed life table person-years lived for men in the age interval x to x + n in year y ( ${}_n{L}_x^y$ ) multiplied by age-, BA-status, and year-specific men’s marital status distributions, we use person-years lived among men aged x to x + n in 1960 ( ${}_n{L}_x^{1960}$ ) for all years from 1880 to 2019 multiplied by age-, BA-status, and year-specific men’s marital status distributions.

Results

Trends in Life Expectancy at Age 15.

Fig. 1 shows trends in male life expectancy at age 15 (e₁₅). Male adult life expectancy has increased and men with a BA have higher life expectancy than those without a BA.^‡ This education difference grew to an 8-y difference in 2019. It is remarkable that men without a BA experienced a decline in life expectancy beginning in 2012. A decline also occurred for the total male population beginning in 2014 but did not for men with a BA. Increases in U.S. life expectancy have been slower in recent decades than those of other countries. American men’s life expectancy lags behind other high-income countries, for example, Britain, Germany, France, and Japan (44).

Fig. 1.

Male life expectancy at age 15 by year and BA status.

Despite the troubling lack of improvement and recent decline in life expectancy among men without a BA, in comparison with earlier eras, men in 2019 had more adult years of life (age 15+) to spend married than in the past (62 y vs. 43 y in 1880). Men with a BA had, in particular, a large number of potential adult years to spend married (68 y).

Change in Marriage and Other Marital Statuses.

The expected number of adult years lived in the five marital statuses we differentiate is determined by 1) changes in mortality (summarized by adult life expectancy in Fig. 1) and 2) marital status trends. Fig. 2 shows trends in the second component.^§ The notable decline in the percent married since 1970 is clear. Compared with 1970 when 71% of men aged 15 and older were married, 49% were married in 2019. Fig. 2 shows that the decline since 1970 is accounted for primarily by increases in the percentage of men never married, followed by cohabitation, and then by divorce/separation (1970 to 2019 percentage point increase: 13, 6, and 5, respectively).^¶ Men with and without a BA alike experienced declines in percentages married, but this decline was substantially greater for non-BA men (SI Appendix, section F).

Fig. 2.

Distribution of marital status for men aged 15+ (age standardized to 2019).

Variation in the intensity and timing of entry and exit from marriage and the other marital statuses underlies these trends. For example, trends in the percentage of men who are never married are the result of variation in first marriage timing and the prevalence of lifelong bachelorhood. Between 1880 and 1940, the percentage of the male population never married remained relatively stable due to offsetting trends of a decline in men’s age at first marriage along with an increase in the prevalence of older bachelors (45). The percentage of men never married hit an all-period low during the Baby Boom era of the 1950s and 1960s both because of early and near universal marriage. Since then, the percentage of men never married has grown due to very high median ages at first marriage (about 30 y of age for men in 2019) and increases in marriage foregone (refs. 45 and 46; authors’ calculations based on U.S. decennial census and ACS data). Although the majority of adults now cohabit prior to marriage (76% of marriages formed in 2015 to 2019), the typical cohabiting union is short-lived and thus, as Fig. 2 shows, only a small proportion of men are in cohabiting unions at any given time (47, 48).

Although divorce rates have declined since the late 1970s (12), the percentage of men divorced/separated has grown likely because of a substantial decline in remarriage rates. Men’s remarriage rates declined from 117 per 1,000 previously married men in 1970 to just 39 per 1,000 in 2017 (49), thus counteracting the effect of declining divorce/separation rates.

Expected Lifetime Years Married and in Other Marital Statuses.

How have changes in mortality and marriage patterns combined to produce expected lifetime years married and in other marital statutes? Fig. 3 shows the composition of men’s adult life expectancy in each of the five marital statuses. The total height of the bars is life expectancy at age 15 among men surviving to age 15 (e₁₅), also presented in Fig. 1. Expected adult lifetime years never married has fluctuated since 1880 but was at a record high of almost 20 y in 2019. The number of expected lifetime years cohabiting has increased, rising from 1.7 y in 1990 to 3.7 in 2019. Years spent in the divorced/separated state have also increased from about 2 y in 1960 to almost 6 y in 2019. Expected lifetime years widowed has stayed relatively constant for men at around 2 y since 1880.

Fig. 3.

Composition of men’s life expectancy at age 15 in each marital status by year.

To home in on trends in the number of years married, Fig. 4A plots expected adult lifetime years married for the total male population, and from 1960, by BA status. Several results stand out. Among all men, expected lifetime years married increased from 27 in 1880, peaking at 38 y in 1960 and 1970 before declining thereafter to 31 y in 2019. Expected lifetime years married in 2019 was at its lowest level for the male population since 1930.^# Thus, despite increasing life expectancy, reductions in marriage have been large enough to completely offset this, cutting 7 y from expected lifetime years married for men since 1960.

Fig. 4.

Men’s lifetime years married and cohabiting (as expected at age 15). (A) Expected years married. (B) Expected years married or cohabiting. (C) Expected years married holding mortality constant at 1960 levels.

Fig. 4A also shows large education differences. Expected lifetime years married in 1960 for men with a BA was about 3 y greater than for those without a BA, but the difference has grown much larger. Remarkably, expected lifetime years married for men with a BA has declined little since the golden age of marriage in the Baby Boom era. In 2019, expected adult lifetime years married for men with a BA was 40 y compared with 41 y in 1960. By contrast, expected lifetime years married for non-BA men declined precipitously from 38 y in 1960 to 27 y in 2019. In 2019, the difference in expected lifetime years married between BA and non-BA men was 12.5 y. Lifetime years married for men without a BA in 2019 was nearly as low as it was for the male population in 1880 when life expectancy was short and marriages were much more likely to end by death rather than divorce.^||,**

Fig. 4B shows that increases in nonmarital cohabitation have offset a large portion of the decline in expected years married for the total population of men and non-BA men, but not all. If we consider cohabiting unions as marriages and examine all coresidential unions (Fig. 4B), expected lifetime years married for men would have declined by 3.4 y (instead of 7) since 1960, thus accounting for slightly over half (52%) of the decline. For BA men, there is an increase in years spent partnered when including cohabitation. Educational differences are nearly as large as for marriage alone (12 vs. 12.5 y).

To what extent do changes in mortality account for these trends? To assess this, we estimate counterfactual trends holding men’s age- and BA-status-specific mortality rates constant at their 1960 levels but allowing marriage patterns to vary as observed. Fig. 4C suggests that the increased longevity of BA men is responsible for a large portion of the observed relative stability in lifetime years married. When we hold mortality constant at its 1960 values, lifetime years married for BA men fall from 41 y in 1960 to 32 years in 2019, rather than the observed 1-y drop. Differences in years married by BA status are also narrower under this counterfactual, a 36% reduction from 12.5 y in the observed data to 8 y. This illustrates how changes in mortality and marriage patterns combine to magnify BA status differences in expected lifetime years married.

Although lifetime years married for men with a BA has been relatively stable since 1960, delayed marriage and to a lesser extent increased divorce/separation and cohabitation mean that these years are a smaller proportion of men’s total lifetime (SI Appendix, section B). The potential outcomes of marriage and cohabitation are generally measured in terms of years of exposure, but it is possible that perceptions of the proportion of one’s life spent in various marital states also affect behavior. Future research should investigate how perceptions and experiences of time horizons affect key outcomes of interest.

Discussion

Our estimates show that men’s expected adult lifetime years married peaked in the United States in 1960 and 1970 and has declined since then to levels comparable to 1930. In addition, very different patterns characterize men with and without a BA. For men with a BA, expected lifetime years married has remained relatively stable since the heyday of marriage in the Baby Boom era, decreasing by just one year (from 41 to 40 y) between 1960 and 2019. By contrast, expected lifetime years married for men without a BA has declined precipitously to 27 y, a low not seen for the male population in the United States since 1880. In 2019, men without a BA could expect to spend 12.5 fewer years of their lives married than men with a BA if the 2019 age-specific marriage and mortality rates remain constant. Cohabitation accounts for a substantial fraction of the decline in expected lifetime years married for men (about half) and BA-status differences when cohabiting unions are included are nearly as large as for marriage alone.

These results tell a different story than past research. Watkins et al. (1) examined trends for women through 1980 and found that, despite much public and academic commentary about changes in marriage and family life in this era (e.g., refs. 50 and 51), expected years married was quite high relative to what had come before, although lower than in the Baby Boom era. They speculated that increased life expectancy itself may have redefined people’s expectations about marriage and family, stating, “It is as if having approached the full potential for family life inherent in low mortality [in the Baby Boom era], subsequent cohorts trimmed the sails sharply” (1:354).

Although perceptions about life expectancy may indeed affect marital behavior and be an important part of the explanation for the trends shown here, we see the evidence as more consistent with an economic inequality and instability explanation. Men with a BA do not appear to have substantially trimmed the sails on lifetime years married since 1960. By contrast, lifetime years married for men without a BA has declined far more than a return to pre-Baby Boom 1940s’ levels. Economic precarity in men’s and women’s lives and the declining economic fortunes of less educated men have made it increasingly difficult to achieve the economic security many people desire to have in marriage (52–54).

Period estimates of expected lifetime years married summarize mortality and marital status conditions in the population in a given year and, like period estimates of life expectancy, are extremely useful as such (32, 33). In addition, past research has shown that period life expectancy measures are good estimates of cohort life expectancy, but are lagged. Goldstein and Wachter (31) estimate that the length of this lag is about 40 y in recent data. If this relationship holds with the added complexity of changes in marriage, then these estimates would be most applicable to the lifetime experiences of men roughly aged 55 in 2019. Going forward, given that younger cohorts with and without a BA alike are marrying later and cohabiting more, we might expect to see declines in lifetime years married as these cohorts age.

Future research should seek to better understand the implications of these shifts by continuing to tease out the (likely changing) causal effects of marriage, cohabitation, and other marital statuses, and the extent to which these effects compound with duration. Regardless of their effects, however, our results show that the combination of changes in mortality and marriage patterns has resulted in high and relatively stable expected lifetime years married for men with a BA and rapidly declining expected lifetime years married for men without a BA. Increasing inequality in both life expectancy and marriage patterns amplify differences in lifetime experiences of coresidential partnerships.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Our code and instructions for generating our results can be accessed on the Open Science Framework (OSF) at https://osf.io/9JSQU/ (60). All data used are publicly available and most are easily downloaded. Users can create their own accounts to access publicly available IPUMS https://www.ipums.org (32) and Human Mortality Database data https://www.mortality.org (31). Mortality data by BA status from 1990 to 2019 are also publicly available and easily downloadable https://www.nber.org/research/data/mortality-data-vital-statistics-nchs-multiple-cause-death-data (44). Data not easily downloaded are provided on OSF.