Older Adults’ Attitudes Toward Cohabitation: Two Decades of Change

Abstract

Objectives:

Our study tracks cohort change in the attitudes of adults aged 50 and older toward cohabitation from 1994 to 2012.

Method:

We used data from the 1994, 2002, and 2012 waves of the General Social Survey to examine the roles of cohort replacement and intracohort change in the trend toward favorable cohabitation attitudes and to examine sociodemographic variation in patterns of support for cohabitation.

Results:

Support for cohabitation accelerated over time with nearly half (46%) of older adults reporting favorable attitudes toward cohabitation in 2012 versus just 20% in 1994. This shift in older adults’ attitudes largely reflected cohort replacement rather than intracohort change. Some of the factors associated with later life cohabitation experience were linked to supportive attitudes.

Discussion:

Cohort succession is fueling the growing acceptance of cohabitation among older adults and coincides with the rapid growth in later life cohabitation that has occurred in recent decades.

Family attitudes have changed dramatically in recent decades, reflecting rising acceptance of new family forms such as cohabitation (Powell, Bolzendahl, Geist, & Steelman, 2010; Thornton & Young-DeMarco, 2001). The growing approval of less traditional family living arrangements dovetails with key shifts in U.S. family patterns that are marked by a retreat from marriage, high levels of divorce, and increases in families formed outside of marriage through cohabitation (Cherlin, 2010; Manning, 2004; Seltzer, 2000; Smock, 2000). These family patterns are evident across the life span, even among middle-aged and older adults aged 50 and older. Fewer older adults are married and rising shares are cohabiting (Brown, Bulanda, & Lee, 2012; Vespa, 2012). The modest uptick in divorce (Kennedy & Ruggles, 2014) is most pronounced among those older than 50 years, for whom there has been a gray divorce revolution (Brown & Lin, 2012). Today, one in three Baby Boomers is single versus just one in five in this age group in 1980 (Lin & Brown, 2012).

In short, there are more unmarried older adults these days, and they increasingly form cohabiting rather than marital unions. The growth in cohabitation among those aged 50 and older has accelerated, tripling from slightly less than 1 million (Brown, Lee, & Bulanda, 2006) to more than 3.3 million persons just from 2000 to 2013 (U.S. Census Bureau, 2014). This trend portends greater acceptance of new family forms such as cohabitation among older adults. But the limited research to date on families in later life has not addressed whether the attitudes of older adults toward cohabitation have changed over time. Nor do we know which older adults hold the most (or least) favorable views toward cohabitation.

Using data from the 1994, 2002, and 2012 General Social Survey (GSS), we examine trends in the cohabitation attitudes of individuals aged 50 and older, investigating the roles of cohort versus period effects. Specifically, we examine the extent to which the growing favorability of cohabitation attitudes is a function of cohort replacement versus intracohort change (Firebaugh, 1992, 1997; Ryder, 1965). Our study addresses whether the factors associated with older adults’ cohabitation (Brown et al., 2006; Chevan, 1996; Hatch, 1995) are also predictive of holding favorable attitudes toward cohabitation later in life and whether these patterns differ by gender. The findings from this study contribute to the emerging literature on intimate relationships in older adulthood by establishing a nearly two-decade time trend of older adults’ attitudes toward cohabitation.

Background

Cohabitation is on the rise, and nowadays, nearly one in four cohabiting couples include at least one member aged 50 or older (U.S. Census Bureau, 2014). The growth in older adults’ cohabitation is not merely an artifact of the aging of the population or the rise in unmarried older adults. Instead, it represents a true increase in the older adults’ cohabitation rate. Expressed as the share cohabiting among those at risk (i.e., unmarrieds), the rate has doubled from 4% of older unmarrieds in 2000 to 8% today (authors’ calculations). At first glance, this trend would seem to foretell a corresponding increase in older adult acceptance of cohabitation in recent decades. Still, most single older adults are not cohabiting. In fact, non-coresidential dating relationships are more common than cohabitation among singles in their 50s and beyond (Brown & Shinohara, 2013). Living apart together (LAT) relationships are another option favored by some older unmarrieds (De Jong Gierveld & Merz, 2013; Duncan & Phillips, 2011). Older singles typically remained unpartnered, forming neither a cohabiting nor a marital union (Brown et al., 2012; Vespa, 2012). For these reasons, population shifts in older adults’ attitudes toward the acceptability of cohabitation may have changed relatively little over the past couple of decades, suggesting that period change in attitudes toward cohabitation may be minimal.

Attitudes toward cohabitation were first tracked in the mid-1970s among high school seniors, whose views have become increasingly favorable over time. Acceptance of cohabitation rose in the 1980s and 1990s among U.S. adults, who increasingly endorsed living together outside of marriage (Thornton & Young-DeMarco, 2001). The suitability of cohabitation as a prelude to marriage or step in the courtship process also rose during this time period, as growing shares of Americans agreed that it is a good idea to live together prior to getting married. Women’s attitudes toward cohabitation consistently have been less favorable than men’s. Nonetheless, by the early 1990s, a majority of women and men alike affirmed that living together was all right, if the couple had plans to marry.

Historically, there was a notable generation gap in support of cohabitation. Drawing on longitudinal data from middle-aged mothers and their young adult children, Thornton and Young-DeMarco (2001) found that although both generations grew more favorable toward cohabitation over time, offspring were twice as likely as mothers to express approval of cohabitation in the early 1990s. Whether this generation gap persists today is an empirical question that the present study aims to answer by investigating the relative contributions of cohort replacement versus intracohort change (Firebaugh, 1992, 1997). Certainly, as cohabitation has become more widespread across the entire population, it is likely that the generation gap has diminished over time, but it may still exist, particularly if cohort replacement is the force underlying the attitudinal shift.

A trend toward acceptance of cohabitation among older adults is evident from qualitative studies that have shown a greater inclination to cohabit versus marry, especially among women (Davidson, 2001; Talbott, 1998; Watson & Stelle, 2011). Cohabitation offers older adults many of the benefits of marriage, namely an intimate, coresident partnership, without the legal entanglements of marriage. Older adults often prefer cohabitation to marriage precisely because it allows them to maintain their autonomy, financially and otherwise. Cohabitation permits individuals to continue to receive Social Security or pension benefits from a former spouse that they would lose were they to remarry. It also helps to protect assets and financial resources from being drained by the partner, facilitating economic transfers across generations. Finally, cohabitation does not appear to carry the same expectations for caregiving as marriage, a salient issue at this stage of the life course, particularly for women (Noel-Miller, 2011). In fact, many older women desire neither cohabitation nor marriage and instead seek male companionship that does not involve either a long-term commitment or coresidence. LAT relationships offer these benefits to older adults and appear to be growing in popularity particularly among older women (De Jong Gierveld & Merz, 2013). In contrast, older single men are much more interested in forming a coresidential union (McWilliams & Barrett, 2014).

Support for cohabitation is likely to be greatest among those who are cohabiting, although this is a small group. Nevertheless, the defining characteristics of older adult cohabitors arguably could be reflective of the factors associated with favorable cohabitation attitudes. Prior research has shown that older cohabitors are a distinctive group, whose demographic characteristics, economic resources, and social ties tend to differ from those of either older marrieds or unpartnereds (Brown et al., 2006; Chevan, 1996; Hatch, 1995). Older cohabitors are disproportionately men, reflecting the imbalanced sex ratio at older ages (Chevan, 1996) and the disinclination of older women to form unions (Talbott, 1998). Most older cohabitors are divorced rather than widowed or never married. They tend to be relatively young compared with either marrieds or unpartnereds. They also are less often White and more often Black or Latino. The economic profile of older cohabitors differs by gender, with cohabiting women more disadvantaged than their married counterparts, whereas among men, cohabitors and marrieds are similar across education, household income, home ownership, and health insurance coverage. Cohabitors typically have weaker social ties than either marrieds or unpartnereds. They are less religious and have fewer friends and family nearby (Brown et al., 2006).

The Present Study

The current investigation uses nationally representative data to establish a nearly two-decade time trend of attitudes toward cohabitation. Our focus is on the attitudes of adults aged 50 and older, a group that is increasingly embracing cohabitation over marriage (Brown et al., 2012; Vespa, 2012). As the Baby Boomers swell the ranks of the aged and growing shares of older adults will be unmarried, we can anticipate a further acceleration in later life cohabitation (Cooney & Dunne, 2001). For these reasons, it is important to identify how the attitudes of older adults toward cohabitation have changed over time.

From a social change perspective, the growth in support for cohabitation may reflect cohort succession, within-cohort change, or some combination of both factors (Firebaugh, 1992; Ryder 1965). Cohort succession, also termed cohort replacement, describes population turnover. Over time, later born cohorts replace earlier born cohorts as the earlier born cohorts die off and the later born cohorts age into later life. For example, if the Baby Boomers are distinctive in terms of their high levels of support for cohabitation, then as they age into older adulthood (i.e., the population turns over and cohorts with less favorable attitudes are replaced), we would expect the attitudes of older adults as a group to be more favorable toward cohabitation. Cohort effects represent stability among a distinct set of individuals (i.e., a cohort) who have shared similar formative experiences that endure over the life course (Alwin & McCammon, 2003; Ryder, 1965). Alternatively, within-cohort or intracohort change occurs when individuals actually experience attitudinal shifts over time, typically due to period effects that reflect the historical context or events that have similar effects across all age groups. Intracohort change is evidenced when individuals within a given cohort express more favorable attitudes toward cohabitation today compared with two decades ago. We investigate the relative contributions of cohort replacement and intracohort change in growing support for cohabitation.

Our study also addresses how the trend in older adults’ attitudes compares with that of younger individuals. Prior research documented a generation gap in attitudes toward cohabitation, but this gap occurred during the early 1990s (Thornton & Young-DeMarco, 2001). A constellation of factors suggest that this gap probably has greatly diminished over time. First, cohabitation has diffused more widely across the population. Second, many of today’s older adults experienced cohabitation themselves at younger ages as members of the Baby Boomer generation that was the first to cohabit in large numbers. Third, even older adults who have not cohabited previously often have indirect experience with cohabitation through their children, who have formed cohabiting unions. Cohabitation among offspring is positively linked to subsequent cohabitation among parents (McClain, 2011). In short, the generation gap in cohabitation attitudes ought to have declined considerably over the past two decades. Our study assesses whether this convergence reflects cohort succession or within-cohort change.

In addition to documenting trends in the cohabitation attitudes of older adults, we also investigate the correlates of holding favorable attitudes toward cohabitation. Drawing on prior literature that has identified the key characteristics associated with cohabitation in later life, we examine whether these same factors are linked to older adult favorability toward cohabitation. Specifically, demographic and economic factors ought to be related to older adults’ attitudes toward cohabitation. Age is negatively associated with cohabitation, indicating that earlier born cohorts ought to be less likely to support cohabitation than later born cohorts. Men are much more likely to cohabit than women and single men also express stronger desires to form unions than do women. Racial and ethnic minorities experience a broader array of living arrangements and are less likely to be married and more likely to be cohabiting than Whites. These patterns suggest Blacks and Hispanics hold more supportive cohabitation attitudes, although the reverse pattern is also possible, given that minorities tend to express relatively conservative attitudes about family life (Powell et al., 2010). Divorced and never-married individuals should hold more favorable attitudes than marrieds and widoweds, who tend to be more traditional. Economic resources may be tied to cohabitation attitudes such that those with higher levels of education and income are more supportive, aligning with patterns obtained for family attitudes more generally. Those with more resources tend to express more liberal attitudes (Powell et al., 2010). We also examine the role of social ties, such as having children and religious attendance, because both of these factors are negatively associated with cohabitation in later life (Brown et al., 2006).

Gender plays a critical role in older adults’ intimate relationships. Older men and women seem to have unique preferences and desires for partnering that may impinge on their attitudes toward cohabitation. Men enjoy a growing pool of eligible partners as they age because men typically partner with women of similar or younger ages, whereas women face a shrinking pool because they usually partner with men older than themselves. Furthermore, the gender asymmetry in later life partnering means that the characteristics of those who partner often differ by gender. For example, cohabiting women appear disadvantaged economically compared with remarried women, but this pattern does not hold for men (Brown et al., 2006). Thus, we test for gender interactions in the full model to determine whether the factors associated with favorable attitudes toward cohabitation differ for men and women.

This study moves the field forward by elucidating the roles of cohort versus period change in older adults’ attitudes toward cohabitation over the past two decades. It illustrates how accepting older adults are of cohabitation these days and which older adults are most likely to be in favor of cohabitation. Understanding the trends in and correlates of cohabitation attitudes will help us to anticipate the magnitude and contours of future growth in later life cohabitation.

Method

Data for this study were pooled from the 1994, 2002, 2012 waves of the GSS, the 3 years in which respondents were asked about their attitudes toward cohabitation. The GSS is a continuous collection of cross-sectional survey data administered by the National Opinion Research Center (Smith, Marsden, Hout, & Kim, 2013). Using face-to-face interviews in the 48 contiguous states, the survey targets nationally representative samples of adults aged 18 and older. A major focus of the GSS is on the opinions, and changes in the opinions, of Americans, which is facilitated by repeating the same or similar questions in several rounds of data collection.

The 3 years of pooled data produced an overall sample size of 7,731 respondents. However, not every respondent in the GSS is asked each question. Of the 7,731 respondents, 3,811 were eliminated because they were not asked their opinion on the acceptability of cohabitation (n = 3,920). An additional 93 respondents were eliminated due to missing data on this dependent variable (n = 3,827). Another eight respondents were eliminated because they were missing on the age variable (n = 3,819). To account for missing data on the control variables (which amounted to just 2% on income and much less on all other variables), we used mean substitution on continuous variables and modal substitution on the categorical variables. This produced a sample size of 3,819, of which 1,465 respondents were aged 50 and older.

Measures

Dependent Variable

In each of the 3 years of data, respondents were asked, “Do you agree or disagree it is alright for a couple to live together without intending to get married?” Cohabitation attitude was measured as a binary variable, with respondents who either strongly agreed or agreed coded as 1 and those who neither agreed nor disagreed, disagreed, or strongly disagreed coded as 0. This coding strategy allows us to track changes in the level of support (or agreement) for cohabitation over time.

Focal Independent Variable

Cohort was captured using a series of dummy variables to classify respondents into 10-year birth cohorts: 1905–1914, 1915–1924, 1925–1934, 1935–1944, 1945–1954, and 1955–1964 (reference). Year of birth was calculated by subtracting the respondent’s age from the survey year.

Demographic Characteristics

Gender was coded with man as 1 and woman as 0. Race was measured as a dichotomous variable with 1 for White and 0 for non-White. There were not sufficient sample sizes of Blacks and Hispanics to permit measurement of these two groups separately. Marital status was a series of binary variables: married (reference), divorced or separated, widowed, and never married. Cohabitation was only measured in 2012 (1 = yes and 0 = no).

Economic Resources

Education was coded as a series of binary variables, reflecting less than a high school education, high school graduate (reference), some college, and college educated. Employment status was captured by three binary variables: full-time employment (reference), part-time employment, and other work, which includes those retired, in school, those looking for work, or temporarily away from their jobs due to illness, vacation, or strike. Income was measured as total family income in the year of interview. The response categories differed by the year of interview (in 1994, the range was under $1,000 to $75,000 or more; in 2002, it was under $1,000 to $110,000 or more; and in 2012, it ranged from under $1,000 to $150,000 or more). Moreover, these figures are not in constant dollars and thus do not account for inflation. To accommodate these issues, we reclassified income into quartiles with roughly 25% of the sample in each of the four income categories each year. The first quartile ranged from under $1,000 to $17,499 in 1994, under $1,000 to $19,999 in 2002, and under $1,000 to $22,499 in 2012. The second quartile included respondents whose reported family income was between $17,500 and $34,999 in 1994, $20,000 and $39,999 in 2002, and $22,500 and $49,999 in 2012. The third quartile included respondents with family incomes ranging from $35,000 to $59,999 in 1994, from $40,000 to $74,999 in 2002, and from $50,000 to $89,999 in 2012. Finally, the fourth quartile included those with incomes of $60,000 and more in 1994, $75,000 and more in 2002, and $90,000 and more in 2012. Respondents who were missing on income were assigned the second quartile, the modal category. A flag was created to mark respondents who were missing on income, with those missing coded as 1 and those not as 0.

Social Ties

Respondents were asked to report the number of biological children they had, and we dichotomized this variable to reflect 0 = childless and 1 = one or more children. Religious services attendance was measured as a continuous variable and responses ranged from 0 = never to 8 = more than once a week.

Analytic Strategy

We began by documenting the trend in older adults’ favorable attitudes toward cohabitation over the 1994–2012 period. Next, we constructed a period-by-cohort table to assess the relative contributions of within-cohort change versus cohort succession for social change in older adults’ cohabitation attitudes. This approach was described in detail by Firebaugh (1992, 1997) and used by Norpoth (1987). Additionally, we estimated levels of support for cohabitation across the other study variables to assess how cohabitation attitudes changed over time among each of the demographic subgroups. Multivariate logistic models were estimated for the full sample to gauge how much of the social change in older adults’ cohabitation attitudes is a function of cohort replacement (i.e., a cohort effect) versus intracohort change (i.e., a period effect). Essentially, we estimated a two-factor model assuming age effects were null given the average age of the population changes little over the 18-year time period (Alwin & McCammon, 2003; Glenn, 2003). Our two mechanisms of interest are changes in the population due to the changing cohort composition (cohort replacement) and the effects of period factors that produce change within cohorts (intracohort change). The logistic models also revealed whether the factors that have been linked to older adults’ cohabitation were predictive of attitudinal support of cohabitation. Additional models were estimated to test for gender interactions. Finally, a model predicting cohabitation attitudes in the 2012 survey round was estimated to evaluate whether the higher likelihood of cohabitation support among later born cohorts was an artifact of current cohabitation (which was only measured in 2012). To correct for the complex sample design of the GSS, descriptive and multivariate analyses were conducted using the svy procedure in Stata.

Results

Descriptive Results: Attitudinal Trends

For the entire adult population, attitudes toward cohabitation have remained stable since 1994, as shown in Figure 1. At all three time points, about 50% of adults agree that living together without intentions to marry is acceptable. For adults aged 18–49 years, the share expressing supportive attitudes toward cohabitation has increased from 52% in 1994 to 64% in 2012. By comparison, the growth in supportive attitudes among Americans aged 50 and older over this time period has been much more dramatic. In 1994, just 20% of older adults expressed favorable attitudes toward cohabitation. By 2002, 30% of older adults were supportive of cohabitation, representing a 50% increase over the 8-year period. In the past decade (i.e., 2002–2012), this share grew by another 50%, bringing the level of support to 46% for today’s middle-aged and older adults. The overall trend for all adults obscures variation by age. Still, the generation gap in cohabitation attitudes that was uncovered in the early 1990s (Thornton & Young-DeMarco, 2001) appears to be closing. How much of the shift in older adults’ attitudes toward cohabitation reflects cohort replacement versus intracohort change?

Figure 1.

Supportive attitudes toward cohabitation by age group and year.

Table 1 shows within-cohort and total change in the percentages of older adults with supportive attitudes toward cohabitation by cohort and period (Firebaugh, 1997). Each row of the table illustrates intracohort change, which has been rather modest averaging just 3% during the 1994–2002 time period and 7% during the 2002–2012 period. By comparison, the total change is larger, as indicated by the columns for each of the three time periods. The totals for each period represent the weighted averages across cohorts during the year. The magnitude of the total change is relatively large, with attitudes shifting by an average of 10% during the 1994–2002 period and 15% during the 2002–2012 period. In short, this period-by-cohort table indicates that the growth in favorable attitudes toward cohabitation among older adults is primarily driven by cohort replacement rather than intracohort change.

Table 1.

Within-Cohort and Total Change in Percentage of Older Adults With Supportive Attitudes Toward Cohabitation, 1994–2012

	1994		2002		2012		Change
Cohort	%	N	%	N	%	N	1994–2002	2002–2012
1905–1914	7.6	49	—	—	—	—	—	—
1915–1924	9.1	131	14.6	38	12.5	12	−5.5	2.1
1925–1934	18.1	140	14.6	104	22.5	60	3.5	−7.9
1935–1944	30.7	182	23.2	130	37.0	126	7.5	−13.8
1945–1954	—	—	51.9	136	54.2	174	—	−2.3
1955–1965	—	—	—	—	51.1	179	—	—
Total (all cohorts)	19.3	502	29.8	408	44.9	551	−10.5	−15.1
Average within-cohort change (weighted by size)							3.1	−7.0

Note: Percentages are weighted to correct for the complex sampling design of the General Social Survey.

In addition to varying across cohorts, trends in the acceptance of cohabitation differ across other demographic characteristics, as shown in Table 2. Attitudes toward cohabitation have become more favorable among both men and women. In 1994, 23% of men and 18% of women agreed or strongly agreed that cohabitation without plans to marry was acceptable. By 2012, 51% of men and 41% of women held this view. In 1994, non-Whites (24%) were more favorable toward cohabitation than were Whites (20%), but today the pattern is reversed with Whites (47%) more supportive than non-Whites (39%). Cohabitation support varies by marital status. The divorced have consistently expressed the most supportive attitudes toward cohabitation over time, with 42% favorable toward cohabitation in 1994 and 65% today. Never marrieds have joined the divorced individuals, with support rising from 24% in 1994 to 65% today. The married and widowed are comparatively less enthusiastic about cohabitation. Just slightly more than one third of both groups reported supportive attitudes in 2012. Not surprisingly, those who are currently cohabiting are highly supportive of cohabitation. In 2012, 88% of cohabitors expressed favorable attitudes toward cohabitation.

Table 2.

Weighted Percentages of Older Adults Who Support Cohabitation by Year (n = 1,465)

Variable	1994 (n = 502)	2002 (n = 413)	2012 (n = 550)
Demographic characteristics
Gender
Women	18.4	27.7	40.9
Men	22.7	33.9	51.1
Race and ethnicity
White	19.5	31.2	47.2
Non-White	24.1	24.7	38.5
Marital status
Cohabiting	—	—	87.5
Married	17.3	29.4	38.8
Divorced	42.4	43.4	64.6
Widowed	12.3	14.9	35.2
Never married	24.2	43.6	64.5
Economic characteristics
Education
Less than high school	18.5	27.5	38.4
High school graduate	21.4	21.8	39.7
Some college	21.1	33.1	49.8
College	18.5	39.8	53.3
Employment
Full-time employment	25.7	46.0	54.6
Part-time employment	38.5	26.3	34.8
Other work	13.8	22.2	41.9
Income
Quartile 1	19.4	22.3	50.5
Quartile 2	22.9	22.9	35.4
Quartile 3	18.3	33.1	52.8
Quartile 4	16.9	48.3	53.5
Social ties
Religious attendance (median = 3)	22.4	25.5	62.5
Children	20.1	28.9	41.8
No children	19.1	42.1	67.1

Note: Analyses are weighted to correct for the complex sampling design of the General Social Survey.

Two decades ago, older adults’ cohabitation attitudes varied little by education, ranging from 18% to 21% in support. Nowadays, there is a distinct education gradient, with higher levels of education corresponding with more widespread support of cohabitation. Only 38% of those who did not finish high school express favorable attitudes toward cohabitation versus 53% of those with a college degree. The pattern by employment has changed course over time. In 1994, the full-time employed (26%) were more favorable toward cohabitation than those engaged in other forms of work (14%) but less favorable than those working part-time (39%). Now, the full-time employed are the most supportive (55%), followed by those performing other work (42%), and finally the part-time employed (35%). There is little variation in cohabitation support by income quartile; growth in favorable attitudes toward cohabitation has occurred among all four quartiles.

Social ties are gauged by religious attendance and having had children. Support for cohabitation nearly tripled (from 22% to 63%) among those reporting the median level of religious attendance. The share of older adults with children who report favorable attitudes toward cohabitation doubled from 20% in 1994 to 42% in 2012. Among childless older adults, support for cohabitation rose more dramatically from 19% in 1994 to 67% in 2012.

Multivariate Results

As shown in Table 3, social change in older adults’ attitudes toward cohabitation predominantly reflects cohort replacement rather than intracohort change, a conclusion consistent with the patterns documented in Table 1. Older cohorts are less likely to support cohabitation than younger cohorts. The odds ratios indicate a consistent rise across subsequent cohorts until the Baby Boom. Early Boomers, born 1945–1954, do not significantly differ from Late Boomers, born 1955 and later, suggesting that either the cohort effect may be diminishing for the youngest older adults or the Boomers are better conceptualized as one large cohort. Period, or year of survey, is not significantly related to the likelihood of supporting cohabitation, indicating that the growth in supportive attitudes has not occurred within cohorts but rather through cohort succession. The association between cohort and cohabitation attitudes is not modified by period (result not shown), underscoring the centrality of cohort replacement.

Table 3.

Logistic Regression Model Predicting Supportive Cohabitation Attitudes (N = 1,465)

	Odds ratio	Confidence interval
Cohort
1905–1914	0.09***	0.03–0.29
1915–1924	0.17***	0.07–0.39
1925–1934	0.31***	0.17–0.57
1935–1944	0.57*	0.35–0.93
1945–1954	1.19	0.74–1.94
1955–1964 (ref)
Year
1994 (ref)
2002	0.84	0.53–1.32
2012	1.21	0.76–1.93
Gender
Women (ref)
Men	1.16	0.86–1.56
Race and ethnicity
White	1.16	0.76–1.77
Non-White (ref)
Marital status
Married (ref)
Divorced	2.56***	1.79–3.65
Widowed	1.50†	0.97–2.31
Never married	1.41	0.74 – 2.70
Education
Less than high school	0.98	0.61–1.55
High school graduate (ref)
Some college	1.26	0.85–1.88
College	1.49*	1.01–2.19
Employment
Full-time employment (ref)
Part-time employment	1.07	0.63–1.84
Other work	0.89	0.64–1.25
Income
Quartile 1 (ref)
Quartile 2	0.86	0.56–1.30
Quartile 3	1.14	0.72–1.83
Quartile 4	1.33	0.78–2.27
Income flag	0.96	0.58–1.59
Religious attendance	0.75***	0.72–0.79
Children	0.70	0.44–1.12
No children (ref)

Notes: Analyses are weighted to correct for the complex sampling design of the General Social Survey.

^† p < .10. *p < .05. **p < .01. ***p < .001.

Few demographic characteristics are linked to cohabitation attitudes in the multivariate model. Neither gender nor race and ethnicity are associated with cohabitation attitudes net of the other variables in the model. Thus, women and men as well as Whites and non-Whites are similarly disposed toward cohabitation. Attitudes toward cohabitation vary by marital status. Relative to married individuals, the odds of supporting cohabitation are 2.5 times higher among divorced individuals and 1.5 times higher among widoweds (p < .10). Never-married individuals do not significantly differ from marrieds in their likelihoods of supporting cohabitation.

Economic factors do not appear to be closely tied to middle-aged and older adults’ attitudes toward cohabitation. Only education is related to attitudes such that the college educated are distinct from those with a high school degree. The former are more likely to support cohabitation than the latter group. Those without a high school diploma or only some college do not differ from their counterparts who completed high school. Nor is either employment or income related to cohabitation attitudes.

Social ties are linked to support for cohabitation. Religious attendance is negatively associated with favorable attitudes toward cohabitation. However, having offspring is not significantly related to cohabitation attitudes, after taking into account other covariates.

Figure 2 depicts the predicted probabilities (derived from the multivariate model shown in Table 3) of reporting supportive attitudes toward cohabitation by 10-year cohort (5-year cohorts yielded a comparable picture, not shown) and across time periods. This figure provides strong visual evidence of a cohort effect, with higher likelihoods of cohabitation support predicted for each successive cohort, except the two latest born cohorts (the Early and Late Boomers) are slightly reversed, with the probabilities roughly .05 higher for the 1945–1954 cohort than the 1955–1964 cohort, perhaps signaling convergence. Note there is very little positive slope in each line, indicating minimal period effects. The predicted probability of holding a supportive attitude toward cohabitation ranges from about .15 for the 1915–1924 cohort to .49 for the 1945–1954 cohort. Similarly, the range in 2012 spans from .19 for the 1915–1924 cohort to .56 for the 1945–1954 cohort.

Figure 2.

Predicted probability of cohabitation support by cohort.

Contrary to our expectation, results are overwhelmingly similar for men and women. Only one significant gender interaction emerges in the full model (interaction results not shown). The association between marital status and attitudes toward cohabitation is modified by gender. Widowed men are much more likely to express supportive attitudes toward cohabitation than are widowed women (odds ratio = 3.46, p < .01). There is no appreciable gender gap in support of cohabitation among those who are divorced, never married, or married.

Finally, we estimate a model predicting cohabitation attitudes in 2012 to examine a more detailed specification of marital status that distinguishes cohabitors from the other unmarried groups. An initial model (not shown) with only the cohort measures included reveals the same pattern documented in Table 3: younger cohorts are more likely to support cohabitation than older cohorts, but the two Boomer cohorts are not distinguishable. As shown in Table 4, the cohort differentials attenuate somewhat with the inclusion of the controls. Older cohorts born 1915–1924 and 1924–1934 are less likely to support cohabitation than are Late Boomers. But younger cohorts born 1935 or later are similarly likely to support cohabitation. They are also the most likely to be currently cohabiting. Indeed, the odds that a cohabiting individual reports favorable attitudes are about 4 times higher than for marrieds. As expected, cohabitors are much more likely to be supportive of cohabitation than marrieds. Other correlates operate similarly to the full model presented in Table 3 with one exception: those with children are much less likely to be supportive of cohabitation.

Table 4.

Logistic Regression Model Predicting Supportive Cohabitation Attitudes in 2012, Net of Cohabitation Experience (N = 550)

	Odds ratio	Confidence interval
Cohort
1915–1924	0.11*	0.01–0.78
1925–1934	0.31*	0.12–0.79
1935–1944	0.81	0.42–1.58
1945–1954	1.03	0.57–1.87
1955–1964 (ref)
Gender
Women (ref)
Men	1.25	0.78–2.02
Race and ethnicity
White	1.09	0.57–2.07
Non-White (ref)
Marital status
Married (ref)
Cohabiting	4.11*	1.30–12.98
Divorced	2.65***	1.51–4.62
Widowed	1.97†	0.91–4.22
Never married	1.43	0.54–3.79
Education
Less than high school	0.79	0.36–1.78
High school graduate (ref)
Some college	1.39	0.69–2.81
College	2.06*	1.09–3.91
Employment
Full-time employment(ref)
Part-time employment	0.58	0.26–1.32
Other work	0.85	0.50–1.42
Income
Quartile 1 (ref)
Quartile 2	0.55	0.29–1.06
Quartile 3	1.31	0.53–3.22
Quartile 4	1.11	0.43–2.89
Income flag	0.75	0.36–1.54
Religious attendance	0.70***	0.66–0.76
Children	0.52*	0.28–0.98
No children (ref)

Notes: Analyses are weighted to correct for the complex sampling design of the General Social Survey.

^† p < .10. *p < .05. **p < .01. ***p < .001.

Discussion

During the past two decades, the attitudes of older adults toward cohabitation have become increasingly favorable. Nearly one half (46%) either agree or strongly agree that a couple living together without intending to marry is acceptable. In 1994, just one in five expressed favorable attitudes toward cohabitation. As we anticipated, the cohabitation attitudes of older adults now more closely mirror those of younger age groups. Ultimately, the generation gap documented in earlier research (Thornton & Young-DeMarco, 2001) is closing.

Indeed, the shift in cohabitation attitudes among older adults over the past two decades appears to reflect cohort succession rather than intracohort or period change. Both the descriptive period-by-cohort table and the multivariate models indicated that later born cohorts are more favorable toward cohabitation than are earlier born cohorts. In fact, period was not significant in the multivariate model net of cohort, meaning that intracohort change plays a negligible role in the growing support for cohabitation among older adults. The predominance of cohort replacement aligns with the growth in cohabitation among successive cohorts that arguably emerged with the Baby Boomers, the first generation to cohabit in large numbers as young adults. Boomers were too young for the 1994 round but Early Boomers were in the 2002 survey and by the 2012 survey both Early and Late Boomers were present, constituting the majority share of the 50+ population. Notably, Early and Late Boomers are similar in their levels of support for cohabitation, either indicating a slowdown in the cohort replacement effect or underscoring the unique experiences of the Boomers, Early and Late alike. In short, cohort replacement appears to be the primary engine behind the rapid growth in support for cohabitation among middle-aged and older adults.

The higher levels of support for cohabitation among younger cohorts coincides with the rapid growth in cohabitation that has occurred for today’s older adults. Our analysis of the 2012 GSS that included a measure of current cohabitation showed that cohort membership continued to be related to attitudes toward cohabitation, although the effect was less pronounced. Later born cohorts (1935–1965) remained more supportive of cohabitation than earlier born cohorts (1915–1934), net of whether they were currently cohabiting. Of course, later born cohorts are more often cohabiting. Unfortunately, with only one time point (2012), we cannot disentangle whether more supportive attitudes prompted higher levels of cohabitation or the reverse. This is an important topic for future research.

Drawing on prior research that has documented the sociodemographic correlates of later life cohabitation, we examined whether and how these same factors are linked to the likelihood that middle-aged and older adults report favorable cohabitation attitudes. Attitudes varied by marital status, with the divorced and widowed especially likely to be favorable toward cohabitation. Support was greater among widowers than among widows. Economic characteristics were largely unrelated to cohabitation attitudes, but social ties played a key role. In particular, religious attendance was negatively associated with agreeing or strongly agreeing that cohabitation is acceptable, and this association held for men and women alike. Contrary to our expectations, there was no significant gender difference in support of cohabitation in the full model. In an initial model that included only cohort, period, and gender (not shown), older men were more likely to support cohabitation than were older women. But this initial gender differential was reduced to nonsignificance with the inclusion of the control for religious attendance (men averaged lower levels of attendance than did women). Similarly, a marginally significant (p < .10) racial gap in an initial model (not shown) was no longer significant with controls included.

This study offers the first look at how older adults think about cohabitation. Notably, we tracked the levels and change in attitudes among adults aged 50 and older across a nearly 20-year time span using national data. At each of the three time points, the identical item was used to measure cohabitation attitudes. We uncovered evidence for cohort succession as a key driver of growth in favorable attitudes toward cohabitation among older adults. The role of intracohort change appears negligible.

Nevertheless, this study also has some limitations. First, there is only one measure of cohabitation attitudes that is available across all three time points. In 1994 and 2002, a second item was asked about cohabitation, but this item was not included in the 2012 GSS survey and thus it was not examined here. Our measure of support for cohabitation is arguably rather stringent because it asks respondents about their views of a couple living together without intending to marry (emphasis added). Some people may find cohabitation acceptable if the couple plans to wed. Thus, our study offers a conservative estimate of older adults’ acceptance of cohabitation. Second, data limitations precluded identifying which respondents were themselves cohabiting during the first two time points. This measure was only available in the 2012 round of the GSS. Our results indicate that cohabitors are more likely to report favorable attitudes toward cohabitation than are married individuals, but cohort continues to be independently associated with attitudes such that the later born are more supportive than the earlier born. Also, the GSS does not capture prior cohabitation experience, precluding a potentially rich analysis of how prior behavior is linked to current attitudes. Nor does the GSS distinguish LAT relationships in its measure of marital status, so these individuals are lumped together with other unmarrieds even though LATs may be similar to cohabitors or even marrieds. Finally, we acknowledge that employment and income are not ideal measures of economic well-being for older adults, but the GSS does not include measures of home ownership or assets.

Cohabitation continues to gain support as evidenced by the nearly two-decade upward trend in attitudes toward living together without plans to marry. Cohort replacement appears to be driving this shift, setting the stage for sustained growth in support for cohabitation among older adults. This growth coupled with an increasing share of unmarried adults portends rising levels of later life cohabitation in the coming years, raising new questions about the determinants and consequences of cohabitation for the health and well-being of older adults.

Funding

This research was supported in part by the Center for Family and Demographic Research, Bowling Green State University, which has core funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R24HD050959).