TIME-VARYING CORRELATES OF ADULT SINGLEHOOD: EDUCATION, WORK, LIVING ARRANGEMENTS, AND MENTAL HEALTH

Abstract

Using the Youth Development Study (n = 504 women and 421 men), we examine how changes in education, work, living arrangements (e.g., living with children, parents, and/or roommates), and mental health predict occasions in adulthood when respondents were single (i.e., not living with a spouse or cohabiting with an unmarried partner) from ages 21 to 38. The odds of singlehood were higher when individuals were living with parents or living with roommates. The odds were lower when living with children, as well as after earning a BA/BS degree. Overall, the correlates shed light on the growing population of single US adults.

Introduction

In the United States (US), the percentage of single adults (i.e., neither married nor cohabiting with an unmarried partner) has risen. Though the percentage of cohabiting adults increased over the past thirty years (from 4% to 9%), the percentage of married adults declined from 67% to 53%, and the percentage of never-married adults nearly doubled from 17% to 33% (Fry & Parker, 2021). In 2022, over two-thirds of young adults ages 18 to 29, and approximately one-third of young adults in their 30s, were single (Juteau, 2022). Research shows that single adults have lower educational attainment and are less likely to be employed than adults who are married or cohabiting (Fry & Parker, 2021; Juteau, 2022). In addition, single adults are more likely to reside with parents and less likely to live with a child than partnered adults (Fry & Parker, 2021). Psychological well-being is also lower when individuals are single compared to when they are partnered, either through cohabitation or marriage (Musick & Bumpass, 2012; Sassler & Lichter, 2020).

Following a life course perspective (Elder et al., 2003; Shanahan, 2000), we used prospective data from the Youth Development Study (YDS) to examine the time-varying correlates of singlehood from young adulthood to adulthood. Beginning in 1988, the YDS has followed a randomly selected panel of 1,010 ninth-grade students in St. Paul, Minnesota for over three decades (Mortimer, 2003). Past year singlehood was assessed over ten waves from ages 21–22 to 37–38, giving us panel data from 504 women and 421 men with 3,988 and 2,902 total observations, respectively. At each wave, respondents completed detailed surveys regarding their education, work, health, and relationships, and these surveys included a life history calendar indicating whom they lived with over the past year (e.g., spouses, partners, children, parents, and roommates). We build upon prior research on the correlates of singlehood in two ways. First, the ten waves of data allowed us to assess how changes in education, work, living arrangements, and mental health predicted occasions in adulthood when respondents were single (i.e., not living with a spouse or cohabiting with an unmarried partner) from young adulthood to adulthood. Second, we used fixed-effects logistic regression models (Allison, 2009), which allowed us to assess whether the socioeconomic disadvantages and mental health risks associated with singlehood in cross-sectional research are observed in longitudinal research after controlling for time-stable unobserved factors. The fixed-effects models also allowed us to assess whether selection processes accounted for any observed links between singlehood and living arrangements, such as when respondents were single and living with children, parents, and/or roommates.

A Life Course Perspective on Singlehood from Young Adulthood to Adulthood

Life course research is guided by five principles (Elder et al., 2003), which stress the importance of transition timing, heterogeneity in life paths, linked lives, historical time and place, and bounded agency. Each of these theoretical principles provided insight into how changes in education, employment, living arrangements, and mental health would be associated with changes in singlehood from early adulthood to adulthood.

Principle 1:

Transition Timing

The “demographically dense” young adult years are characterized by rapid social role transitions in school, work, and family (Schoon & Silbereisen, 2009; Shanahan, 2000). Whereas the transition to adulthood was once marked by the completion of full-time school and the acquisition of full-time work, as well as residential independence from parents, marriage, and parenthood, the timing and sequencing of school, work, and family transitions have become increasing delayed and disorderly (Amato et al., 2007; Jager et al., 2022; Schoon & Silbereisen, 2009; Settersten & Ray, 2010). Nowadays, young adults are returning to full-time school after working full-time (Vuolo et al., 2014), returning to their parent’s home after full-time schooling (Houle & Warner, 2017), and having children outside of marriage (Livingston, 2018). Moreover, the once classical social role markers of adulthood are occurring later (Settersten & Ray, 2010); as schooling becomes prolonged, young people reside with their parents longer due to high housing costs and student debt, and the onset of careers, romantic partnerships, and parenthood is postponed. Research shows that these changes are consequential for adult identity, with those not achieving these life course transitions less likely to feel like adults and more likely to feel behind the progress of their peers (Eliason et al., 2015).

According to the life course principle of timing, the impact of social role transitions on singlehood will depend on when and if they occur in a person’s life. Drawing on this principle, we expect that school, work, and residence transitions will impact transitions to and from singlehood, even after controlling for age effects on these associations. For instance, cross-sectional research from the American Community Survey revealed that 26% of single men and 33% of single women in 2019 had completed at least a BA degree, compared to 37% of partnered men and 43% of partnered women (Fry & Parker, 2021). However, these cross-sectional associations overlook the role of attending college in imparting values that encourage singlehood, as college-going single adults may be more likely to transition to cohabitation or marriage after completing their degree (Rindfuss et al., 1980). The odds of singlehood will thus be higher when individuals attend school than when they are not attending school. However, the transition from student to non-student may also have long-term effects on singlehood, as implied by cross-sectional research (Fry & Parker, 2021), in that the odds of singlehood will be lower if respondents had completed a BA degree.

Principle 2:

Heterogeneity in Life Paths

Trajectories of school, work, family, and residence capture changes in social role transitions as individuals move through the life course (Elder et al., 2003). For example, a family trajectory could involve a series of transitions in social roles and changes in living arrangements during the early adult years, such as leaving the parental home, cohabiting, being single (i.e., not living with a spouse or unmarried partner), and then getting married. Similarly, school and work trajectories involve transitions from being a student to not being a student, and transitions from working to not working, as well as capture the timing, intensity (full- vs. part-time; long vs. short duration), and success (complete degree vs. dropout) of these role transitions. Life paths capture the interdependence of these trajectories over time. Educational and work trajectories will have reciprocal effects as people age (Staff & Mortimer, 2007). Family trajectories will affect, and be affected by, trajectories of school and work (Sassler & Lichter, 2020). According to the life course principle of life paths, education, employment, family, and residence trajectories are interdependent, leading to heterogeneity in life paths. Changes in school, work, and residence will affect changes in relationships as young people move through adulthood.

This interdependence among the trajectories may explain why single adults have lower educational and wage attainments than married or cohabiting adults (Fry & Parker, 2021; Juteau, 2022). A college degree is likely to be a strong correlate of singlehood given that: 1) a college degree is often needed to secure a stable full-time job in the contemporary economy (Mortimer et al., 2008); 2) those without college degrees are increasingly likely to experience union dissolution and single parenthood (McLanahan, 2004); and 3) those with college degrees are more likely to maintain supportive unions with partners who also graduated from college (Breen & Salazar, 2011; Mirowsky & Ross, 2003). Earning a college degree is likely to reduce the odds of singlehood by increasing the likelihood of full-time work and a stable partnership. Building on this principle, we also expect that the odds of singlehood will be lower when respondents are in full-time employment compared to when they are not, even after controlling for BA degree attainment and other time-varying correlates.

Principle 3:

Linked Lives

The life course principle of linked lives states that an individual’s life path is interwoven with the role transitions and trajectories of significant others. This dependence on others is most apparent in family trajectories, as transitions from singlehood to cohabitation and marriage involve the formation of a union. However, adult singlehood is often still dependent on others. In 2019, approximately 31% of single men and 24% of single women lived with their parents, compared to 2% of those partnered (Fry & Parker, 2021). In addition, 32% of single women and 8% of single men lived with a child in 2019, compared to approximately 60% of those who were partnered (Fry & Parker, 2021). Single adults are also likely to share residence with a roommate, especially in young adulthood. In 2016, over half of single young adults in the US lived with a roommate when they were not residing with a child or parent (Schweizer & Payne, 2018).

This principle points to changes in living arrangements with parents, children, and roommates as key predictors of singlehood. We expect that singlehood will be more likely when adults reside with parents (especially for men), more likely when they live with roommates, and less likely when they reside with children (especially for men). When assessing how linked lives affect transitions in and out of singlehood, it is important to control for school and work trajectories. For instance, young adults without college degrees are more likely to live with their parents (Fry, 2015). Thus, earning a BA degree will likely reduce the odds of singlehood because it increases the odds of full-time work, a partnership, and economic and residential independence from parents.

Research shows that poor mental health is also a correlate of singlehood (Musick & Bumpass, 2012; Sassler & Lichter, 2020). For instance, singlehood is associated with indicators of mental health such as stress, specifically stress associated with loneliness (Ta et al., 2017). Depressive symptoms are also associated with singlehood, especially for men across adulthood, while only in early adulthood for women (Grundström et al., 2021). However, early adult singlehood can also lessen the effect of adolescent dating on later depressive symptoms, especially for women (Olson & Crosnoe, 2017). Consistent with the principle of linked lives, research has shown that the health correlates of singlehood depend on the overall household structure, indicating that there are both health advantages and disadvantages when singlehood includes others in one’s residence (Hughes & Waite, 2002). At the beginning of the COVID-19 pandemic, about half of young adults who lived alone reported anxiety and depression symptoms, likely due to social isolation (File & Marlay, 2021). Thus, we expect that heightened depressed mood will increase the odds of singlehood, even after controlling for whether the individual is living with others.

Principle 4.

Historical Time and Place

In addition to the influence of significant others (i.e., parents, children, roommates), an individual’s life path is affected by historical time and place. Changes in the timing and ordering of social role transitions have reshaped family life among contemporary adults. For instance, Barroso et al. (2020) showed that in 2019, approximately 30% of Millennials ages 23 to 38 years old lived with a spouse and their child or children, compared to 40% of adults who were ages 23 to 38 years old in 2003 (Generation X), 46% of Baby Boomers (ages 23 to 38 in 1987), and 70% of the Silent Generation (ages 23 to 38 in 1968). The cohort of Millennials also had the highest proportion who were single while at the same time living with a child(ren), a parent(s), and/or other family members (40%), noticeably higher than the previous cohort (Generation X = 30%).

The data we used in this study come from one cohort of youth from one US city (St. Paul, Minnesota) assessed at one historical time (from 1995 to 2011). The panel consists of young people born in the early 1970s (1972–73), members of Generation X, who transitioned to adulthood in a relatively favorable economic climate of the 1990s and at a time when college costs were lower than in succeeding cohorts. However, the advantage of studying this one cohort is the ability to observe trajectories through young adulthood when role transitions were occurring rapidly and when adult singlehood was becoming more common. It is also important to note that approximately 76% of the YDS sample is White, 9% Black, 6% Hispanic, 4% Asian, and 5% other races/ethnicities, reflecting St. Paul at that historical time. The primarily White sample makes it difficult to assess potential differences in singlehood patterns by race/ethnicity. For instance, 59% of Black adults were single in the US in 2019, compared to 38% of Hispanic adults, 33% of White adults, and 29% of Asian adults (Fry & Parker, 2021).

Principle 5:

Bounded Agency

In life course research, individuals have agency in directing how their lives unfold (Hitlin & Johnson, 2015). Consistent with the principle of agency, sociological studies of status attainment show that educational expectations and occupational aspirations in early adolescence are robust predictors of long-term educational attainment and career acquisition, even after controlling for family socioeconomic background (Sewell et al., 1969). Moreover, childbearing preferences strongly predict family formation behaviors and subsequent fertility (Hagewen & Morgan, 2005; Schoen et al., 1999). Nevertheless, at the same time, this agency can be bounded by demographic background, the educational and occupational attainment of their parents, the expectations of significant others, and the historical time and place they live (HitlinJohnson, 2015; Schoen et al., 1999; Sewell et al., 1969).

There is reason to expect that some of the well-known correlates of singlehood, such as low educational attainment, no full-time work, residing with parents, and poor mental health, might partly be driven by individual differences established in adolescence. For instance, though expectations for marriage and children remain high among youth, using a life course framework, youth who do not plan to cohabit or marry as adults are more likely to be single in adulthood. Furthermore, as some longitudinal research has shown that poor mental health is a precursor of subsequent singlehood (Horwitz et al., 1996), high depressive affect in adolescence may increase the odds of singlehood in adulthood. We would also expect young adult singlehood to be more prevalent among youth from lower SES backgrounds, and among Black and Hispanic youth (Brown et al., 2020; Wu, 2017). When following this principle, isolating the time-varying correlates of singlehood requires controlling for sources of bounded agency as well as other potential selection influences.

Current Study

Guided by the principles of life course research, in this study we used fixed-effects logistic regression models and longitudinal data from a cohort of 504 women and 421 men to assess how over-time changes in education, work, living arrangements, and mental health predict occasions in adulthood when respondents were single versus living with a spouse or cohabiting with an unmarried partner from ages 21 to 38.

Method

Participants

In 1988, the YDS surveyed a random sample of 1,010 ninth-grade students enrolled in the St. Paul, Minnesota public school district (Mortimer, 2003). The youth’s parents were also surveyed in 1988. During high school, this sample of youth was administered yearly in-school surveys, with 932 retained by age 17–18. Each study participant was followed longitudinally through surveys (via paper or online) from 1992 to 2019, encompassing 16 waves of data collection following high school. This research was approved by the Human Research Protection Program (IRB ID: 9103S03585) at the University of Minnesota. Informed consent was obtained from all research participants.

For these analyses, we used data from ten waves collected prospectively from 1995 to 2011 that included relevant study variables: wave 8 (age 21–22); wave 11 (age 25–26); wave 12 (age 26–27), wave 13 (age 28–29); wave 14 (age 29–30); wave 15 (30–31); wave 16 (age 31–32), wave 17 (33–34); wave 18 (age 35–36), and wave 19 (age 37–38). Regarding panel retention, of the original 1,010 respondents, 780 were retained by age 21–22, and 653 by age 37–38. Previous studies show that the retained panel by age 21–22 was similar to the initial panel in mental health, socioeconomic background, school achievement, substance use, and family structure, though the percentage of White youth and females was higher in the retained sample (Mortimer, 2003; Staff & Mortimer, 2007).

Measures

Outcome: Past Year Singlehood

At each of the ten waves, respondents completed a life history calendar indicating who they lived with during each month of the prior year (e.g., spouses, partners, children, parents, and roommates). From these life history calendars, we first created our time-varying outcome. Singlehood is coded 1 if respondents indicated that they were single during all months over the past year (i.e., not residing with a spouse or partner), and coded 0 otherwise. Since our analyses are based on yearly and not monthly changes in our outcome and predictor variables, we followed a more stringent view of singlehood (i.e., stably single in all months over the past year) to avoid coding respondents as single if they were married or cohabiting during part of the year.

Time-Varying Predictors

Time-varying predictors of singlehood included changes in education, employment, living arrangements, and mental health from ages 21 to 38. Respondents indicated at each wave whether they had attended school in the past year (1 = no school attendance), and whether they had attained a BA/BS degree or higher (1 = yes). Based on data from the life history calendars, we included a time-varying measure indicating whether they worked full-time all 12 months over the past year (1 = yes). To capture changes in living arrangements, we included three time-varying measures assessing whether they resided with their child (1 = yes), their parents (1 = yes), and roommates (1 = yes) during the entire past year. The time-varying indicator of depressive affect (Ware et al., 1979) was derived from four items indicating how much of the time in the past month respondents were under strain, stress, or pressure, felt downhearted and blue, felt depressed, or felt in low or very low spirits. The item responses, ranging on five-point scales from “none of the time” to “all of the time,” were averaged at each wave (alpha ranged from .83 at age 21–22 to .88 at age 37–38).

Methods

We estimated a series of fixed-effects logistic regression models using the “xtlogit, fe” program in Stata 18 (StataCorp, 2023), with bootstrap standard errors (50 replications). A key advantage of these panel models is that individuals serve as their control, as each respondent is compared to themselves on the predictors to assess whether they are more likely to be single at times when the predictors are different (Allison, 2009). Thus, the effects of time-varying changes in education, work, living arrangements, and mental health on singlehood shown in these models are net of all time-stable confounders, both observed and unobserved. Another strength of our modeling strategy is that YDS respondents were included in any year that they contributed data, giving us panel data from 504 women and 421 men (followed over 3,988 and 2,902 observations, respectively).

Results

Descriptives

Table 1 shows descriptive statistics, by sex, based on the pooled sample. As shown in Table 1, women were single in 48% of observations, and men were single in about 49% of occasions. These pooled statistics mask some sex differences in singlehood by age. As shown in Figure 1, 82% of women were single, defined as not living with a spouse or partner at age 21–22. By age 37–38, the proportion single in the past year had dropped to 35%. The percentage of men single at age 21–22 was higher (87%), and then declined to 32% by age 37–38.

FIGURE 1 Percentage Single by Sex and Age.

TABLE 1 Descriptive Statistics by Sex (% or Mean and SD)

Outcome	Women	Men
Past year singlehood	48%	49%
Time-varying Predictors
Education
Not attending school	72%	76%
Earned BA/BS degree	29%	27%
Employment
Working full-time	53%	69%
Living Arrangements
Living with children	55%	36%
Living with parents	8%	14%
Living with roommates	5%	9%
Mental Health
Depressive affect	2.44	2.30
	(.78)	(.76)
Sample size	504 (3,988 observations)	421 (2,902 observations)

Table 1 also shows descriptive statistics for the time-varying predictors, separately for women and men in the pooled sample. Regarding education, women and men reported not attending school for most observations (72% for women and 76% for men), and women held at least a BA/BS degree in 29% of observations and men for 27% of observations. Time spent in employment and living arrangements varied for women and men. Women worked full-time in 53% of observations, compared to 69% for men. Women spent more of the observations living with children than men (55% for women vs. 36% for men), whereas men spent more time living with parents (14% for men vs. 8% for women) and living with roommates (9% for men vs. 5% for women). The descriptive results also show that depressive affect was higher for women than men on average over the ten waves.

Table 2 shows descriptive statistics of the predictor variables by sex and age, revealing considerable overtime changes in each predictor variable. For all predictors, the most significant changes occurred in young adulthood, during a time of rapid social role transitions. Between the ages of 21–22 and 25–26, we see sizable increases in the percentage of women and men who were not attending school, had earned a BA degree, were working full-time, and resided with children. The percentage of respondents who lived with parents also declined, and women, but not men, experienced a slight decline in the percentage who resided with roommates. Mean levels of depressive affect also declined from ages 21–22 to 25–26.

TABLE 2 Descriptive Statistics of Predictor Variables by Sex and Age

	Not attending school		Earned BA/BS degree		Working full-time		Living with children		Living with parents		Living with roommates		Depressive affect
	%		%		%		%		%		%		mean (sd)
Age	W	M	W	M	W	M	W	M	W	M	W	M	W	M
21–22	44	47	5	4	25	31	23	7	21	33	10	14	2.70 (.73)	2.60 (.74)
25–26	73	74	28	25	59	68	37	16	11	18	8	14	2.45 (.85)	2.19 (.77)
26–27	71	71	29	29	61	75	44	25	8	13	7	13	2.45 (.68)	2.32 (.68)
28–29	68	75	29	29	55	76	53	30	5	12	5	11	2.49 (.74)	2.37 (.73)
29–30	76	77	31	30	59	77	60	40	6	12	7	9	2.55 (.76)	2.35 (.72)
30–31	77	81	31	30	51	72	60	41	6	12	5	6	2.38 (.79)	2.18 (.73)
31–32	80	80	33	32	55	74	65	45	5	12	4	7	2.34 (.80)	2.17 (.73)
33–34	80	86	34	31	53	76	68	51	4	7	1	5	2.43 (.80)	2.31 (.82)
35–36	79	82	35	33	58	75	70	52	5	7	3	4	2.30 (.77)	2.25 (.81)
37–38	77	87	35	33	58	69	75	60	6	8	1	6	2.31 (.78)	2.22 (.79)

W = women; M = men; sd = standard deviation

Sample size = 504 women (3,988 observations); 421 men (2,902 observations)

By age 37–38, 77% of women and 87% of men were not attending school, and approximately one-third of respondents had earned at least a BA/BS degree. Approximately 58% of women and 69% of men worked full-time. Regarding living arrangements, 75% of women and 60% of men lived with children, 6% of women and 8% of men lived with parents, and only 1% of women and 6% of men lived with roommates. Depressive affect, on average, was approximately one-half of a standard deviation lower at age 37–38 compared to age 21–22.

Results from Fixed Effects Logistic Regression Models

Table 3 shows two sets of estimates from fixed-effects logistic regression models predicting singlehood, separately for women and men. The within-person effects shown assess whether changes in the time-varying predictors are associated with a change in the outcome, controlling for age and all other time-stable (static) characteristics of the individual. Approximately 31% of individuals did not show a change in singlehood over the ten observations, and thus do not contribute to these analyses, giving us an average of 8.4 waves of data for women and 7.8 waves for women.

TABLE 3 Fixed-Effects Logistic Regression Predicting Singlehood by Sex

	Women				Men
	logit		Bootstrap SE	OR	logit		Bootstrap SE	OR
Education
Not attending school	−.626	***	(.151)	.535	.232		(.260)	1.261
Earned BA/BS degree	−.852	*	(.360)	.426	−1.334	*	(.633)	.263
Employment
Working full-time	.062		(.163)	1.064	−.733	**	(.238)	.480
Living Arrangements
Living with children	−1.297	***	(.222)	.273	−3.162	***	(.337)	.042
Living with parents	1.816	***	(.330)	6.149	2.346	**	(.737)	10.443
Living with roommates	1.858	***	(.393)	6.411	1.458	***	(.384)	4.297
Mental Health
Depressive affect	.244	**	(.091)	1.276	.017		(.142)	1.017
Age (vs 21–22)
25–26	−.579	*	(.234)	.561	−.760	*	(.358)	.468
26–27	−.918	***	(.226)	.400	−1.202	**	(.352)	.300
28–29	−1.456	***	(.226)	.233	−1.773	***	(.427)	.170
29–30	−1.859	***	(.239)	.156	−2.496	***	(.506)	.082
30–31	−1.465	***	(.260)	.231	−2.159	***	(.467)	.115
31–32	−1.394	***	(.298)	.248	−2.297	***	(.475)	.101
33–34	−1.232	***	(.306)	.292	−2.450	***	(.455)	.086
35–36	−1.483	***	(.288)	.227	−2.444	***	(.454)	.087
37–38	−1.615	***	(.341)	.199	−2.417	***	(.549)	.089
	N = 369 (3,106 observations)				N = 267 (2,078 observations)

*** p < .001, ** p < .01, * p < .05

For women, the odds of singlehood are 46% lower when they are not attending school versus when they are students (OR = .54), and 57% lower when they hold a BA/BS degree versus when they do not (OR = .43). For men, though school attendance is not significantly related to singlehood, earning a BA/BS degree reduces the odds of singlehood by 74% (OR = .26).

Results also show that the effects of full-time employment on singlehood vary by sex. The odds of singlehood are 52% lower when men are working full-time (OR = .48), whereas for women the odds are close to 1 and statistically non-significant (OR = 1.06).

Who was living with the respondent at each wave also has substantial effects on singlehood. For women, the odds of singlehood are 6.2 times higher when residing with a parent (OR = 6.2), 6.4 times higher when living with a roommate (OR = 6.4), and 73% lower when they were residing with a child (OR = .27). The effects of parents and children on singlehood are of higher magnitude for men, as the odds of singlehood are over 10 times higher when residing with a parent (OR = 10.4), and 96% lower when residing with a child (OR = .04). The odds of singlehood are 4.3 times higher when men live with a roommate (OR = 4.3), which is not as strong as the association for women (OR = 6.4).

The within-person estimates also reveal that the odds of singlehood are 28% higher in waves when women are 1 unit higher on depressive affect (OR = 1.28). For men, the odds of singlehood based on changes in depressive affect are close to 1 and statistically non-significant (OR = 1.02). Finally, age has a powerful effect on singlehood for women and men, as the odds of singlehood are lower at older ages compared to ages 21–22.

Table 4 shows predicted probabilities of singlehood based on changes in education and employment, and in living arrangements, separately for women and men. The predicted values were derived from the fixed-effects logistic regression estimates in Table 3, and the table also includes the percentage of observations from each role combination. The probability of singlehood was 25% for women and 34% for men when they did not earn a BA/BS degree or hold a full-time job. The probability dropped to 13% when men held a full-time job (but did not earn a BA/BS degree) and declined slightly for women (to 22%). For both women and men, the probability of singlehood dropped to the low teens when they earned a BA/BS degree (but did not have a full-time job). The lowest predicted probability of singlehood was when women and men earned a BA/BS degree and worked full-time, reducing the probability to 11% for women and 2% for men.

TABLE 4 Predicted Probabilities of Singlehood by Sex (Based on Estimates from Table 3)

Changes in Education and Employment		Women		Men
Earned BA degree	Working full-time	% of observations	Probability of Singlehood	% of observations	Probability of Singlehood
No	No	35.4%	24.6%	25.3%	33.9%
No	Yes	35.8%	21.8%	47.4%	13.4%
Yes	No	11.4%	13.5%	6.1%	10.4%
Yes	Yes	17.4%	11.5%	21.2%	2.4%
		100%		100%
Changes in Living Arrangements
Living with children	Living with parents	% of observations	Probability of Singlehood	% of observations	Probability of Singlehood
No	No	39.4%	29.7%	50.9%	17.3%
No	Yes	5.3%	76.7%	12.9%	68.2%
Yes	No	52.7%	8.1%	35.3%	0.4%
Yes	Yes	2.6%	42.5%	0.9%	5.1%
		100%		100%

Table 4 also illustrates how different combinations of living arrangements change the probability of singlehood. The probability was especially high when women and men lived with their parents and not with children (77% and 68%, respectively), and especially low when respondents lived with children but not parents (8% for women and 0.4% for men). The probability of singlehood was 42% when women lived with children and parents, whereas for men, it was 5%, though this role combination was rare for men in this sample (occurring on less than 1% of occasions).

Discussion

Research based on data from the US Census has shown considerable demographic shifts in singlehood across generations, as the percentage of adults living with a spouse or partner has grown from 29% in 1990 to 38% in 2019 (Fry & Parker, 2021). Drawing from a life course perspective, and taking advantage of prospective longitudinal survey data, with multiple waves including rich life history calendars, we considered how changes in education, work, living arrangements, and mental health from ages 21 to 38 were associated with transitions to and from singlehood. To further confidence that these time-varying associations were not set in motion much earlier in the life course (i.e., principle of bounded agency), we used fixed-effects models to control for time-stable sources of spuriousness, such as sociodemographic background as well as individual differences in adolescent expectations and mental health. There are three key findings from these analyses.

First, we expected and found significant social role changes in school, work, family, and residence during this demographically dense transition period (Schoon & Silbereisen, 2009; Shanahan, 2000). Following studies of singlehood based on nationally-representative data from the US Census (Fry & Parker, 2021; Juteau, 2022), our analyses focused on a relatively narrow definition of singlehood (i.e., those neither married nor cohabiting with an unmarried partner) and thus did not exclude respondents who were engaged, couples who were living apart, and those who were dating (Beckmeyer & Jamison, 2023; Lavender‐Stott et al., 2023). Nonetheless, respondents reported being single on nearly half of the observed occasions. They showed a dramatic decline in prevalence from age 21 to 38 (from 82% to 35% for women, and from 87% to 32% for men, as shown in Figure 1). Consistent with the principle of transition timing, dramatic changes in prevalence were also observed among the predictor variables, highlighting the considerable heterogeneity in school, work, family, and residence trajectories, despite being from just one cohort drawn from one US city at one historical point.

Second, we found evidence for heterogeneity in life paths, as changes in school, work, family, and residence were associated with changes in singlehood, and some of these associations were different for women and men. For instance, women, but not men, had lower odds of singlehood when they were not attending school versus when they were students, even after accounting for BA/BS degree attainment reducing the odds of singlehood for both women and men. This finding may be because women in this sample spent more time in school, on average, than men, especially in their mid to late 30s, and singlehood can facilitate educational attainment (Rindfuss et al., 1980). Results also showed that the association between full-time employment and singlehood varied by sex, wherein the odds of singlehood were lower when men worked full-time, but not when women worked full-time. Interdependence among the school and work trajectories may account for this sex difference, especially given the increasing likelihood of singlehood among women without college degrees (McLanahan, 2004).

Consistent with the principle of linked lives, the effects of living arrangements on the likelihood of being single were large for the within-person effects measuring the effect of change. The positive effect of living with a parent on singlehood may reflect the perceived incompatibility of partnership (married or not) and residential dependence for adult identity (Eliason et al., 2015), although living with a roommate appears to represent a similar liminal state. The effect of living with a child on singlehood is very likely due to the outsized probability of cohabiting or being married among parents. Importantly, even though single adults were not living with a spouse or partner, their lives were still connected with others, as children, parents, and roommates were strong predictors of singlehood.

Third, it is important to note that the associations shown here are likely bi-directional. For instance, our results show that the odds of singlehood were higher on occasions when women had elevated depressive affect compared to times when it was lower. Though research has shown that poor mental health is a correlate of singlehood (Grundström et al., 2021; Musick & Bumpass, 2012; Sassler & Lichter, 2020), we remain cautious in interpreting depressive affect as a cause of singlehood, as there may be bi-directional associations. Individuals with high depressive affect may be more likely to select into singlehood (Horwitz et al., 1996), and being single could also increase depressed mood. In addition, though our fixed-effects models control for time-stable factors, unobserved time-varying influences could also be at play, as impairments in physical and social functioning or changes in anxiety may account for over time individual variation in both singlehood and depressive affect.

Limitations

We note limitations to our analyses. First, our data represent one cohort of youth from one US city assessed at one historical time, and such historical time and place represents a key component of life course theory. Research using cohorts from different locations and generations can assess how time and place shape the correlates of singlehood shown here. Longitudinal samples from larger, recent, nationally representative cohorts will allow researchers to assess potential differences in patterns by race/ethnicity and account for demographic shifts that have occurred across cohorts. We encourage future studies that can attend to different cohorts and racial/ethnic diversity in singlehood experiences. Second, our measure of singlehood is based on whether respondents were living with a spouse or romantic partner; we do not know if they were dating or engaged when they were single. Furthermore, the prevalence of singlehood would have likely been higher had our yearly measure of singlehood been less stringent (i.e., single in all months of the prior year). Finally, the order of influence for our time-varying predictors may likely be reciprocal; for example, depressed mood could be both a cause and an effect of singlehood. Future research should test the influence of heterogeneous categories of singlehood on physical and mental health.

Conclusions

In this study, singlehood represented occasions when individuals were not living with a spouse or romantic partner. Drawing from principles of life course theory, we found that singlehood depended on whether the respondent was living with significant others, as singlehood often co-occurred with children, parents, and roommates. Singlehood also involved significant transitions in school and work, as well as changes in mental health, and these correlates were observed even after control for selection influences (via fixed-effects models). Moving forward, researchers should continue to unpack the heterogeneity of living without a romantic partner in adulthood using longitudinal data from cohorts from different locations and generations.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

De-identified data from the Youth Development Study are available through the Inter-University Consortium for Political and Social Research, University of Michigan: https://www.icpsr.umich.edu/web/ICPSR/studies/24881.

Additional information

Funding

The Youth Development Study was supported by grants “Work Experience and Mental Health: A Panel Study of Youth,” from the National Institute of Child Health and Human Development (HD44138) and the National Institute of Mental Health (MH42843).