Abstract
Measuring childlessness is complicated by the increasing complexity of family structure. Using data from the 2014 Health and Retirement Study, in this research note we compared three definitions of childlessness: (1) respondent never fathered/gave birth to a child, (2) respondent had no children who were living and in contact, and (3) respondent and spouse/partner had no children or stepchildren who were living and in contact. Results showed that the prevalence of childlessness among Americans aged 55 or older ranged from 9.2% to 13.6% depending on which definition was used. The association between select individual characteristics (gender and marital status) and the likelihood of childlessness, as well as the association between childlessness and loneliness and living arrangements, also varied depending on how childlessness was defined. Therefore, how we define childlessness can affect our understanding of its prevalence, correlates, and relationships with well-being. Future research on childlessness should carefully consider the choice of definition and its implications for research and policy discussions.
Introduction
Adult children provide important support for older parents and help facilitate their social engagement (Bookman and Kimbrel 2011; Kendig et al. 2007). However, trends toward later marriage, more lifelong singlehood, and lower fertility in recent decades have led to a growing population of Americans without children (Aughinbaugh et al. 2013; Dykstra 2009; Hayford 2013; Livingston and Cohn 2010). Understanding the prevalence and characteristics of older Americans without children is essential for developing strategies to address their health and social care needs.
Yet measuring childlessness is complicated by the growing complexity of family structure. High levels of marital disruption have resulted in a tripling of the number of remarried Americans between 1960 and 2013 (Livingston 2014). As a result, many individuals without biological children may have stepchildren, whereas some biological children may become estranged from divorced or repartnered parents (Kalmijn 2015; van der Pas and van Tilburg 2010). The frequency of parent–child contact and the extent of intergenerational support also vary substantially (Dykstra and Fokkema 2011), and older parents may outlive their children and become “elder orphans” (Carney et al. 2016).
Thus, the concept of parenthood/childlessness reflects a continuum rather than a simple dichotomy (Albertini and Kohli 2017). Prior studies have measured childlessness using various definitions, ranging from “biological infecundity” to lack of any “quasi-parental ties” (e.g., stepchildren) and differing in whether they account for contact/vital status of the child (Albertini and Kohli 2017; Bures et al. 2009; Koropeckyj-Cox and Call 2007; Margolis and Verdery 2017; Valerio et al. 2021). Depending on the definition used, the size of the childless population, its characteristics, and its health outcomes may differ in meaningful ways. This can have important implications for policy discussions seeking to address the needs of this population.
This study aims to provide new evidence to encourage critical thinking about the measurement of childlessness. To this end, we applied three different definitions of childlessness, ranging from a restrictive biological definition to a more expansive social definition, to a national sample of middle-aged and older Americans. To illustrate how different definitions may affect our understanding of the childless population, we compared the prevalence of childlessness, correlates of childlessness, and distributions of loneliness and living arrangements among the childless population using each definition. Because the social context and meanings of childlessness have changed substantially over time and the effects of parenthood and children can vary at different life stages (Umberson et al. 2010), we included individuals spanning middle age through late life.
Methods
We used data from the Health and Retirement Study (HRS)—a longitudinal panel survey of Americans conducted every two years since 1992. To construct a nationally representative sample of middle-aged and older Americans, we included respondents aged 55 or older. This is the age range for which HRS analysis weights are available for both community-dwelling and institutionalized individuals. We analyzed data from the 2014 wave of the HRS survey, including 17,183 (weighted N = 85.9 million) respondents born in or before 1959. The necessary variables were taken from the RAND HRS Longitudinal File and Family Data File, which contain cleaned and constructed variables compiled from the HRS, as well as the RAND HRS Fat File and HRS Tracker File. Analysis of respondents' loneliness was limited to a random subsample of 6,843 (weighted N = 84.4 million) respondents who completed the 2014 HRS “leave-behind” self-administered questionnaire (Smith et al. 2017). All analyses accounted for the complex sample design of HRS to generate nationally representative estimates.
We considered three alternative definitions of childlessness: (1) respondent never fathered/gave birth to a child (biological definition), (2) respondent had no children who were living and in contact (functional definition), and (3) respondent and spouse/partner had no children or stepchildren who were living and in contact (social definition). Similarities and differences across the three definitions, as well as the variables used to construct each, are detailed in Table 1. For each definition, respondents who had at least one child meeting the stated criteria for children were coded as being a parent. Those with unknown status on the relevant variables or questionable linkage of child information across waves that precluded definitive determination were coded as having unknown childlessness status.
For each of the three definitions, we calculated the proportion of respondents who were childless. We also compared the prevalence of childlessness by key sociodemographic characteristics (age, gender, race/ethnicity, educational attainment, income, wealth, and marital status) using chi-square tests and logistic regressions adjusting for age. To account for economies of scale in consumption, income and wealth were measured using an equivalence scale, that is, by dividing the sum of respondent's and spouse's income or wealth by the square root of two if married or partnered (Congressional Budget Office 2018).
We measured loneliness and living arrangements for each respondent because they are particularly relevant correlates of well-being and social support for childless older adults. Loneliness was measured using an abbreviated, 11-item Revised UCLA Loneliness Scale (Cronbach's alpha = .87) comprising questions on feelings about lacking companionship, being left out, isolation from others, being in tune with others, feeling alone, having someone to talk to, having someone to turn to, having someone who understands, having someone you feel close to, feeling part of a group of friends, and having a lot in common with friends (Hawkley et al. 2005; Hughes et al. 2004; Lee and Cagle 2017; Russell et al. 1980; Smith et al. 2017). Each item was scored on the basis of the frequency of the feelings (1 = often, 2 = some of the time, 3 = hardly ever or never). The overall score (range, 1–3) is the mean value of the 11 items, with negative items reverse-coded such that a higher score indicates greater loneliness. Living arrangement was classified into four mutually exclusive categories: living alone, living with spouse (regardless of whether there were others coresiding in the household), living with others (without spouse), or living in a nursing home. To make these living arrangement categories consistent between the childless and parents, we treated coresiding children as “others.” The loneliness scores of the childless and of parents were compared using t tests, as well as linear regressions that incrementally adjusted for respondents' key sociodemographic characteristics. In these regressions, the coefficient estimate on the childless (vs. parents) indicator reflects the adjusted mean difference in their loneliness score. The distribution of living arrangements between the childless and parents was compared using chi-square tests, as well as logistic regressions that incrementally adjusted for respondents' key characteristics.
Results
After applying the biological definition, the prevalence of childlessness was 13.6% in 2014, about 11% higher than for the functional definition (12.3%) and 48% higher than for the social definition (9.2%) (Table 2). Under all three definitions, childlessness was lowest for individuals aged 75–79 in 2014 and was progressively higher among those who were older or younger (Table 3).
The prevalence of childlessness differed by sociodemographic characteristics. When using the functional definition, the prevalence was higher among non-Hispanic individuals (e.g., non-Hispanic Whites at 13.1% vs. Hispanics at 6.8%), individuals with higher educational attainment (e.g., more than high school at 14.7% vs. less than high school at 7.9%), and individuals who had never married (e.g., never married at 69.4% vs. married/partnered at 7.9%). The prevalence of childlessness demonstrated a U-shaped relationship regarding wealth: 14.6% and 13.5% in the lowest and highest quartiles of wealth, respectively, compared with 11.1% and 10.6% in the middle-low and middle-high quartiles, respectively. These sociodemographic differences remained significant after adjusting for respondents' age. However, differences in the prevalence of childlessness by gender were no longer significant after adjusting for age.
Some of these relationships between sociodemographic characteristics and the prevalence of childlessness were stronger when using the more restrictive biological definition or the more expansive social definition. For instance, after adjusting for age, women were less likely than men to be childless when using the biological definition (odds ratio [OR] = 0.88; p = .04), although there was no significant gender difference when using the functional definition. Compared with respondents who were married or partnered, after adjusting for age, those who had never married had an OR of 26.35 (p < .001) for being childless when using the functional definition, but an OR of 54.58 (p < .001) when using the social definition.
Shifting now to the well-being of childless Americans, childless respondents had greater loneliness, compared to those with children, when using the functional and social definitions, but not when using the biological definition (Table 4). The difference in loneliness between parents and the childless was mostly explained by respondents' socioeconomic status when using the social definition, but was mostly explained by marital status when using the functional definition.
Under all three definitions, the childless were more likely than parents to live alone rather than with a spouse. However, the magnitude of such differences was greater when the social definition was used. In comparing the childless with parents regarding the likelihood of living alone, the adjusted OR was 5.11 (p < .001) when using the functional definition and 10.31 (p < .001) when using the social definition.
Discussion
Using nationally representative data, we described the prevalence and characteristics of childless middle-aged and older Americans. More importantly, we demonstrated how decisions about the definition of childlessness can affect our understanding of the prevalence and correlates of childlessness and its relationship with loneliness and living arrangements.
Depending on the definition used, the estimated percentage of middle-aged and older Americans without children ranged from 9.2% (social definition) to 13.6% (biological definition). After applying an even broader definition of having no living children (including biological, adopted, or stepchildren and regardless of whether the child was in contact), a recent study reported that 7.5% of Americans aged 50 or older were childless (Quashie et al. 2021). Other research has explored alternative definitions of parenthood/childlessness that consider the availability of the circumstances of biological, adopted, and foster children and stepchildren, as well as of children (e.g., vital status, frequency of contact, and proximity of residence) (Albertini and Kohli 2017; Bures and Koropeckyj-Cox 2009). Depending on how childlessness is defined, clearly there can be considerable differences in the estimated size of this population.
Childless older adults are thought to have more limited social networks and support and to be more susceptible to isolation and loneliness, compared with those who have children (Hansen 2012). However, we found greater loneliness among the childless only when using the functional or social definition of childlessness. Simple biologically based definitions (e.g., ever having biological children) may not accurately reflect the availability of support to older parents as children may die or become estranged (Albertini and Kohli 2017). Conversely, defining children too broadly (e.g., including any stepchildren) may overstate the support available to older parents, as stepchildren often provide less assistance than biological children and the level of support provided by stepchildren depends on the quality of their relationship with stepparents and reciprocal help received from stepparents (Clawson and Ganong 2002; Kalbarczyk 2021; Pezzin et al. 2008). Whether the definition of childlessness affects gender differences in the association between childlessness and well-being, as shown in some research (Nelson et al. 2013), also warrants close attention in future research.
The different life experiences and family circumstances that lead to childlessness can have distinct implications for older adults' need for long-term care services and nonfamily support. For instance, older adults without biological children may have quite different needs for long-term care services depending on whether they have adopted children or stepchildren with whom they are close. Yet inadequate scrutiny and clarity in the definition of childlessness in prior research limit our knowledge about this population, hindering our ability to appropriately deploy resources and tailored interventions to address their needs.
The design of future data collection and research on childlessness should consider the varying definitions and their implications. To enhance cross-study comparisons of findings and evaluation of generalizability, individual studies should clearly define how childlessness is measured, explain whether and why some children are excluded, and discuss the potential impact on findings and their interpretation. Moreover, the planning of future surveys should consider the pertinent aspects of parenthood/childlessness and ensure that such data elements (e.g., various types of children and vital/contact status of each child) are adequately captured. Some standardization of data collection methods and reporting quality will likely facilitate future research in this area.
We also note, however, that our results are in some cases consistent regardless of the definition of childlessness used. For instance, we found a consistently lower prevalence of childlessness among individuals aged 75–79 in 2014 (i.e., born in 1935–1939) compared with others who were older or younger. Although our analysis was limited to a cross-sectional assessment and cannot distinguish age and cohort effects, this finding is consistent with prior research demonstrating important generational differences in childbearing among U.S. women—with the total fertility rate peaking among women born in 1935 (averaging 3.0 children per woman) and steadily decreasing to 2.0 per woman among those born in 1960 (Hayford 2013; Kirmeyer and Hamilton 2011). We also found a consistently lower prevalence of childlessness among Hispanic individuals, a higher prevalence among individuals with greater educational attainment, and a U-shaped relationship between an individual's wealth and their likelihood of childlessness. These features of the sociodemographic composition of the childless can inform tailored health and social policies.
Despite these unique contributions, our study lacked information on respondents' reason(s) for childlessness. Prior research has shown that different paths to childlessness (e.g., voluntary childless, outliving all children, losing contact with children) are associated with different patterns of financial and nonfinancial support that a person receives or provides (Albertini and Kohli 2017). Likewise, older adults who choose not to have children have greater psychological well-being and feelings of autonomy and environmental mastery (Jeffries and Konnert 2002), whereas involuntary childlessness is associated with worse psychological well-being (Huijts et al. 2013; Tanaka and Johnson 2016). Research that further incorporates such granularity into the definition of childlessness will provide additional insights. Moreover, we focused on measuring childlessness. We did not fully assess other relevant concepts, such as estrangement, although our functional and social definitions of childlessness excluded children not in contact. We also recognize that our focus on a single wave of the HRS data precluded us from differentiating age, cohort, and period effects. Additional research using longitudinal data to examine age, cohort, and period differences in the prevalence of childlessness and whether these relationships vary by the definition of childlessness will be an important area for future investigation.
This study provided much needed data on the prevalence and characteristics of childless middle-aged and older Americans. The findings demonstrated that our understanding of the prevalence and correlates of childlessness, as well as the well-being of the childless, differed depending on how childlessness was defined. Clearer conceptualization and more precise measurement of the parenthood/childlessness status continuum are needed when designing future research and survey studies.
Acknowledgments
This work was supported by the National Institute on Aging at the National Institutes of Health (grant R01AG049716; PI: Jersey Liang). Preliminary results were presented at the annual meeting of the Gerontological Society of America in November 2020.