Marriage, Kids, and the Picket Fence? Household Type and Wealth among U.S. Households, 1989 to 2019

: Evidence on how parenthood affects household wealth in the United States has been inconclusive, partially because previous studies have decontextualized parenthood from gender, marital, and relationship status. Yet, insights from economic sociology suggest that wealth-related behaviors are shaped by the intersection of identities, not by a binary classiﬁcation of parental status. We examine net worth by the intersection of gender, parental, and relationship status during a period of increasing wealth inequality and family diversiﬁcation. Using data from the Survey of Consumer Finances from 1989 through 2019, we show that aggregate comparisons between parents and non-parents mask substantial wealth variation across nine household types. Despite changing social selection into marriage and parenthood, married parents consistently held a wealth advantage over demographically similar adults in other household types. Married parents’ wealth advantage descriptively arises from homeownership, perhaps because the combined spousal and parental identities are normatively and culturally associated with homeownership

O NE of the striking features of the wealth distribution in the United States is the wide gulf between households with and without children.In 2019, median wealth among households with resident children younger than 18 years was roughly half the level among households without a resident child (Gibson-Davis and Hill 2021).This pattern has long held: since the late 1980s, households with children have had median wealth levels that were half as large as the average for the total population of households (Gibson-Davis and Percheski 2018).On the surface, these descriptive statistics seem to suggest that becoming a parent may carry a wealth penalty.
But simple comparisons of the wealth of households with and without children belie the theoretical and empirical complexity of how parenthood affects wealth.Whether parenthood increases or diminishes wealth is theoretically ambiguous because of potentially offsetting changes to income, savings patterns, debt accrual, and asset acquisition.Having a child may motivate parents to buy a home, switch to a job that offers higher income, invest in stocks, or save money to provide an inheritance for their children.Yet parenthood may also lead parents to decrease their labor supply, spend down their savings, and increase their debts, especially to secure housing in neighborhoods with higher-quality schools or to finance their children's education.Reflecting this theoretical ambiguity, past work has provided inconsistent evidence as to how parenthood is associated with wealth (Conley and Ryvicker 2004;Maroto 2018, Maroto andAylsworth 2017;Painter and Shafer 2011;Schmidt and Sevak 2006;Yamokoski and Keister 2006).
In this study, we argue that the connection between wealth and parenthood can be better understood by applying foundational insights from economic sociology about the expressive and relational work of financial behaviors (Bandelj 2020;Granovetter 1985;Zelizer 2000).Individuals' economic transactions and consumptive behavior reflect individuals' social identities, cultural norms, and intimate ties (Chang 2010;Epp and Price 2008;Zelizer 2013), albeit imperfectly given constraints posed by gender and racial bias and discriminatory institutions.Viewed through this lens, a single dimension of adult status, such as a binary indicator of parenthood, is unlikely to adequately explain much variation in household wealth and related financial behaviors.We posit that a household head's family-related identities shape their decisions around assets, debts, and expenditures in complex ways, and thus a contextualized approach-which considers parental status, marital status, union type, and gender simultaneously-is necessary to better understand how family characteristics pattern household wealth in the U.S. population.
Furthermore, we argue that associations between parenthood and wealth may vary over time because of temporal shifts in parenting norms and demographic selection into parenthood (Gibson-Davis and Rackin 2014; Smock and Schwartz 2020).Over the last few decades, raising children in the United States has become an increasingly resource-intensive activity with heightened expectations for what parents should provide for their children (Kornrich and Furstenberg 2013).Meanwhile, childlessness has become more pervasive and socially acceptable, and the family structures in which adults raise their children have shifted (Hayford 2013;Kornrich and Furstenberg 2013;Smock and Schwartz 2020).Because demographic shifts in parental composition and evolving expectations of parenthood have differed by socioeconomic status, associations between parenthood and wealth may vary by social class (Gibson-Davis and Rackin 2014;Lareau 2003).Thus, a focus on mean differences in wealth between parents and non-parents at a point in time may miss considerable heterogeneity in the relationship between parenthood and wealth.
Previous work has provided point-in-time mean comparisons in wealth while largely overlooking the multiple spheres that adults inhabit across parental status, marital status, union type, and gender.To address these limitations, we analyze how parenthood, contextualized by gender, marital, and relationship status, may differ in its association with wealth over time.Data come from the 1989 through 2019 waves of the Survey of Consumer Finances (SCF) for 41,625 households with a household head younger than 65 years.We employ a richer classification of households than heretofore has been considered, as adults are classified by their parental status and by the intersection of their gender, marital status, and union type, resulting in nine subcategories of household type: five categories of parent households (married couples, cohabiting couples, never-married mothers, divorced mothers, single fathers) and four categories of non-parent households (married couples, cohabiting couples, single men, and single women).The SCF is ideally suited for our analysis, as it is the only U.S. data set on wealth that captures the top of the wealth distribution and provides comparable nationally representative wealth data over three decades.Thus, with these data we can compare associations between contextualized parenthood and wealth at the bottom, middle, and top of the wealth distribution, and we can investigate how the association between parenthood and wealth has changed from the late 1980s to the late 2010s.Finally, we investigate several possible mechanisms for differences in wealth by household type, with a focus on the role of homeownership.
Our work offers several contributions.Our study is the first to demonstrate that aggregate differences in net worth between parents and non-parents mask considerable variation by the intersection of family-related identities.Specifically, we find that the combination of marriage and parenthood is associated with the largest wealth advantages; this advantage is largest for households near the bottom and middle of the wealth distribution.Second, our study is the first to show that the associations between household type and wealth have remained stable over time, which is surprising given changing selection into parenthood (Hayford 2013) and changes in the normative expectations of parental investments (Bandelj and Grigoryeva 2021;Schneider, Hastings, and LaBriola 2018).Finally, our study is the first to clearly identify homeownership as the primary mechanism accounting for wealth differences by household type.Homeownership is not only the primary means by which families build wealth (Gibson-Davis and Hill 2021;Wolff 2017); it also has strong normative significance in identity building for American families (Gibson-Davis 2009).

Background
Previous research on how parenthood is related to wealth among U.S. households has provided mixed evidence as to the nature of the association.Studies have suggested that parenthood is associated with higher wealth (Painter and Shafer 2011;Yamokoski and Keister 2006), lower wealth (Conley and Ryvicker 2004;Schmidt and Sevak 2006), or no difference in wealth (Maroto and Aylsworth 2017;Ozawa and Lee 2006).Other studies have noted that the association between parenthood and wealth may vary depending on the age of the child or the relative wealth level of the household (Maroto 2018;Schmidt and Sevak 2006).
We argue that these conflicting findings could be potentially reconciled if three shortcomings were addressed: recognizing the intersectionality of parenthood with other adult identities, addressing wealth heterogeneity across time and by place in the wealth distribution, and considering the mechanisms that may descriptively account for the association.We discuss each of these shortcomings in turn.
First, previous research on net worth has decontextualized parental status from other family identities and from household structure (but see Yamokoski and  Keister 2006 1 ).Net worth has been modeled as the additive product of separate identities, with discrete categorical variables for parental status, marital status, and gender.Previous research on wealth and single dimensions of family characteristics finds that married-couple parents have more wealth than unmarried single parents (Grinstein-Weiss et al. 2008;Hao 1996;Lusardi, Cossa, and Krupka 2001), that unmarried fathers have more wealth than unmarried mothers (Chang 2010;Grinstein-Weiss et al. 2008;Hao 1996;Yamokoski and Keister 2006), and that gender differences in wealth among young, unmarried childless adults are either small or nonexistent (Schmidt and Sevak 2006;Yamokoski and Keister 2006).Only one extant study (Yamokoski and Keister 2006) has investigated how parenthood moderates the association between marriage and wealth, and no research to date has analyzed whether wealth levels among cohabiting couples vary by parental status.Thus, although the existing research is informative about wealth differences between narrowly defined comparison groups, this decontextualized approach leaves many questions unanswered about the intersection of wealth and family demography.
Importantly, the decontextualized approach does not reflect sociological understandings of the importance of intersecting family identities or of how most people make financial decisions (Bandelj 2020;Zelizer 2000).Insights from economic sociology and sociology of consumption emphasize that individuals' economic transactions and consumption are deeply influenced by their social identities and relationships (Granovetter 1985;Granovetter and Swedberg 2018).Individuals engage in consumptive activities-such as buying a home, taking out student loans, or investing in the stock market-that are consistent with their perceived identities and intimate ties (Epp and Price 2008;Zelizer 2013).Thus, married parents are likely to make wealth-related decisions that are based on their identities as married adults and as parents (Love 2010), just as an unpartnered man with no children will invest his assets and acquire debts that resonate with his experience as a single, childless individual.Moreover, these identities intersect with social and gender norms as to how adults respond to and govern economic choices (Chang 2010).For example, men have more wealth than women (Maroto and Aylsworth 2017), but these differences are heavily informed by women's identities as mothers and by the wealth-depleting effects of motherhood (Conley and Ryvicker 2004;Ruel and Hauser 2013).
The second shortcoming in the existing literature on parenthood and wealth gaps is that such gaps have been treated as static, and previous research has not examined changes over time or across cohorts.In general, studies have considered either a single birth cohort, such as Baby Boomers (Maroto 2018;Yamokoski and Keister 2006), or a short time period, such as a single year (Grinstein-Weiss et al. 2008).It is unclear whether the inconsistent findings on the associations between parenthood and wealth reflect the aforementioned problem of decontextualization or important secular changes in the associations between parenthood and wealth.
Temporal or cohort changes seem likely, as contemporary parents have adopted a more time and resource-intensive approach to child-rearing (Kornrich and Furstenberg 2013;Sayer, Bianchi, and Robinson 2004).Lareau (2003) has described this approach as "concerted cultivation," or the belief that successful child raising requires a concentrated and deliberate amount of a parent's temporal, monetary, and psychological resources.The rise of concerted cultivation has covaried with increasingly child-centric parental economic activity (Kalil et al. 2016;Ramey and Ramey 2009;Schneider et al. 2018).These changes pertain not just to income, but also to wealth: parents in the mid-2010s, relative to those in the late 1980s, employed more of their assets to support their children's development (Bandelj and Grigoryeva 2021).Notably, though, child-based expenditures differ by social class, marital status, and gender (Hastings and Schneider 2021;Ishizuka 2019;Lareau 2003), likely leading to differences over time between parents and non-parents and by household structure differences among parents.
Additionally, temporal changes have occurred because of changes in sociodemographic selection into parenthood (Hayford 2013) and increased family diversification over the past four decades (Smock and Schwartz 2020).Variation over time in net worth differences by parental status could have occurred because of compositional changes in the population of parents and non-parents and/or changes in the behaviors of these same groups.Compositional changes seem particularly likely, given changes in demographic selection into both childlessness and married parenthood (Gibson-Davis and Rackin 2014;Hayford 2013).Although women with a bachelor's degree are still more likely to be childless, the educational gradient in fertility has attenuated over time (Hayford 2013;Martinez, Daniels, and Febo-Vazquez 2018).At the same time, women with college degrees account for increasing shares of married births (Gibson-Davis and Rackin 2014).Cohabitation has also emerged as a socially acceptable household arrangement and context for childbearing (Sassler and Miller 2017), likely confounding comparisons of single and married adults' wealth status over time.
The third limitation of existing research on wealth and family characteristics is that we do not know which mechanisms underlie wealth differences by parental status and household type.Beyond differences in income and household head demographic characteristics, previous research on wealth has highlighted several potential mechanisms but has not provided strong evidence that any particular behaviors or net worth components account for a substantial share of the wealth gap by parental status or household type. 2 Candidate explanations include differences in financial risk taking and portfolio composition (Love 2010), inheritance receipt (Yamokoski and Keister 2006), business ownership (Grinstein-Weiss et al. 2008), unsecured debt holdings (Grinstein-Weiss et al. 2008), in vivo transfers (Hao 1996), and homeownership (Mulder 2013).
Differences in homeownership are arguably the strongest candidate explanation for parental status and household type wealth gaps.Scholarship dating back to the mid-1950s (Rossi 1955) has recognized the intertwined nature of homeownership and significant life events, such as the transition to parenthood (Mulder and Hooimeijer 1999;Mulder and Lauster 2010).Homeownership-with its connotations of security, stability, and safety-aligns closely with the normative and social expectations of parenthood and further demonstrates that an adult can take on the financial burdens of parenthood (Dupuis and Thorns 1998;Zelizer 2013).Parenthood and homeownership are perceived as complementary life course events that can mark the transition to adulthood (Clark, Deurloo, and Dieleman 1994;Mulder 2006;Mulder and Billari 2010).Although no extant study of which we are aware has compared homeownership by parental status, we expect such differences to be large, given the importance many U.S. parents place on obtaining housing in neighborhoods with access to high-quality schools and other child-centric amenities (Bayer, Ferreira, and McMillan 2007).
For many decades, the transition to parenthood was nearly synonymous with the transition to marriage, and the normative and cultural ties between marriage, parenthood, and homeownership are strong (Gibson-Davis 2009;Mulder and Lauster 2010;Mulder 2013;Smits and Mulder 2008).Couples often buy a home soon after marrying.Many likely buy a house in anticipation of parenthood; the vast majority of married couples have children within the first five years of marriage (Hayford, Guzzo, and Smock 2014).Married couples also encounter societal expectations that married couples should own their homes (Edin and Kefalas 2005).
Although births to cohabiting couples have increased substantially over the past 30 years (Smock and Schwartz 2020), normative expectations of married and cohabiting couples still differ in some domains (Sassler and Miller 2017), and the homeownership norm does not seem to adhere to cohabiting couples (Mulder 2006;Ozawa and Lee 2006).Although cohabiting parents may also desire the benefits of homeownership for their children, joint property ownership is riskier for cohabitors, who lack a legal process for union dissolution.Additionally, cohabitors in the United States cannot file their federal taxes jointly, which affects their access to tax breaks for homeownership.Similarly, single parents may experience different social norms than married parents around homeownership.Moreover, single parenthood is often a short-lived status, as many single parents re-partner (Smock and Schwartz 2020), and unpartnered adults often prioritize flexibility in housing decisions (Mulder 2013).
The fraying of the link between marriage and parenthood also means that an increasing share of married women remain childless (Hayford et al. 2014).Do married couples without minor children have similar homeownership rates to their counterparts who are parenting?To our knowledge, this question has not been previously investigated.We expect that, all else equal, married childless couples will have lower homeownership rates than married parents because they are not concerned with providing children with residential stability and access to high-quality schools and other child-centric amenities.
To summarize, assessing the association between parenthood and wealth has been hampered because prior work has not considered the intersectionality of parenthood with other adult identities, has overlooked variation by time or place in the wealth distribution, and has not examined how wealth-related behaviors contribute to net worth differences.Although this complexity belies simple a priori hypotheses, we postulate that wealth differences between parents and non-parents likely mask considerable heterogeneity across gender, marital status, and union type.Married couples likely have wealth advantages over other household types, but how marriage moderates the association between parenthood and wealth is unclear.Changes over time in the association are likely.Given increasing investments in children by middle-and upper-middle-class parents (Lareau 2003;Schneider et al. 2018), we expect widening gaps by parental status at the top of the wealth distribution, although changing demographic selection into parenthood may intersect with changing investment norms in unforeseen ways.Homeownership is the most likely mechanism to explain variance between parents and non-parents, although the extent to which it can descriptively account for wealth disparities is unclear.

Method Data
Data come from the Survey of Consumer Finances (SCF), a cross-sectional survey of U.S. families conducted by the Federal Reserve approximately every three years.The SCF is nationally representative of households in the United States when weighted and includes an oversample of high-income households.It is considered the best source of U.S. data on wealth and assets (Kennickell 2008).We use data from the 1989, 1992, 1995, 1998, 2001, 2004, 2007, 2010, 2013, 2016, and 2019 waves and restrict our sample to households with a head younger than 65 years (N = 41, 625).All dollar amounts were converted to 2019 dollars using the Consumer Price Index.

Variables
Net worth.Net worth measures the sum total of a household's fungible assets minus debts.Assets include the values of real estate and housing, bank deposits, stocks, financial securities, and trust fund equity.Debts covers the amount owed in several categories, including housing debt (both primary and secondary), consumer debt, educational loans (for 1992 on), and other debt.We followed Wolff (2016) and excluded from our definition of net worth the value of vehicles, whose resale value is far less than the consumption value to the family.We also excluded the value of future pension (i.e., 401(k) or IRA) and Social Security income.(Robustness checks in which the net worth measure included vehicles and the value of future pensions and Social Security yielded a similar pattern of results.)To account for economies of scale, we adjusted net worth by the square root of household size.(Results, however, are largely the same if we do not adjust for household size.)Household size changed only modestly between 1989 and 2019.
Following Killewald, Pfeffer, and Schachner (2017), we model a household's wealth as the household's percentile rank in a weighted net worth distribution.Using the percentile rank of wealth, rather than a log of wealth, means that all households, even if they have negative wealth, can be included in the same regression model (approximately 20 percent of the sample has negative net worth).We calculate percentile net worth separately for each survey year, and therefore the net worth level corresponding to a specific percentile fluctuates across years.An advantage of net worth percentiles calculated within each year is that this measure is not affected by year-to-year changes in the shape of the wealth distribution or the length of the tails of the distribution.
Contextualized parental status.Household heads were categorized into two broad categories-parents and non-parents-and then, within each of these categories, by the gender, marital status, and union type of the household head.Any household head who reported having a resident or nonresident biological, adoptive, or stepchild of any age was classified as a parent.Households headed by a parent were divided into five subcategories of parent households: married parents, cohabiting parents, never-married mothers, divorced mothers, and single fathers (including both never-married and divorced).Non-parent (or childless adult) households consist of four subcategories: married couples, cohabiting couples, single men sociological science | www.sociologicalscience.com (never-married or divorced), and single women (never-married or divorced).Current relationship status superseded past relationship status, so that a divorced mother who is currently cohabiting was classified as cohabiting.Same-sex couples could not be identified as a separate category.Widowed household heads were relatively rare, so we did not divide this group by parental status or gender; instead, we classified widowed heads as a single category, and we do not show results for these families.Altogether, we present results for nine mutually exclusive household type categories.
Demographic characteristics.emographiccovariates in regression models include the following characteristics of the household head: age (linear and squared), race and ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, other race or ethnicity), and education (less than high school, high school, some college, bachelor's degree, or postgraduate degree).
Household income.Income included all earnings, rents, alimony or child support, and all other types of income.Income was divided by the square root of household size and logged and was included in models as a linear and squared term.
Mechanisms.We examined five factors related to wealth building.Attitude toward investment risk was a categorical variable for whether the respondent reported taking average risk, above average risk, or no risk on investments and financial matters.Savings habits measured whether the respondent regularly sets money aside for savings, irregularly saves, or never saves.Inheritance receipt was a dichotomous variable for ever having received an inheritance.The presence of debts was measured through three indicator variables for having credit card debt, educational debt, or vehicle debt.Asset holdings were captured by three dichotomous indicators for homeownership, owning liquid retirement accounts, and owning stocks and financial securities.

Descriptive Statistics
Across years, 73 percent of households were headed by a parent, 24 percent were headed by a childless adult, and three percent were headed by a widowed household head (parent or non-parent).The predominant household type was married parents (47 percent of the sample), followed by divorced mothers (10 percent) and single childless men (eight percent).The other household types-cohabiting parents, never-married mothers, single fathers, cohabiting childless couples, and single childless women-each accounted for seven percent of the sample or less.(There are few observations for cohabiting couples and never-married mothers prior to 2001.)Over time, married-parent households decreased as a percentage of households (from 71 percent in 1989 to 58 percent in 2019), whereas the share of cohabiting parent and cohabiting childless couples approximately tripled.Other household types either experienced modest changes or remained relatively static.Descriptive statistics for the full sample and by parental status are presented in Table S1 of the online supplement.Sample sizes by household type and year are available in Table S2 of the online supplement.

Analyses
We first describe the wealth distribution by parental status and household type.For the broad categories of parents and non-parents, and for each of the more detailed household types, we estimated median wealth (adjusted for household size) and the mean percentile net worth, using SCF data from 2016 and 2019.We then used ordinary least squares regression models to predict percentile net worth by parental status and household type.We specify three model specifications: model 1 includes only year fixed effects, model 2 adds household head demographic characteristics, and model 3 adds household income.To facilitate comparisons across the range of household types, we present predicted wealth percentiles from model 3.
In the second part of our analysis, we used unconditional quantile regression to examine whether parental status and household type differences are similar throughout the wealth distribution.(For more on unconditional quantile regression models, see Firpo, Fortin, and Lemieux [2009] and Killewald and Bearak [2014].)We present results from models estimated near the 25th, 50th, and 75th percentiles of the distribution.Models included controls for household head demographic characteristics, household income, and year covariates.For never-married mothers, 87 percent of observations are below the 50th percentile and only three percent are above the 75th percentile, so we caution against interpretation of the estimates for this household type at the 50th and 75th percentile.We also note that the number of observations for cohabiting couples (parents and non-parents) is low across the wealth distribution in 1989 and 1992.
Our third set of analyses examined differences in wealth gaps by household type over time.We graphed the average net worth percentile of each household type by year, and we modeled percentile net worth as a function of year interacted with household type.Because of small sample sizes, we did not use all nine household type categories but instead collapsed all unmarried-parent households into one category and all unmarried childless households into another category.These categories were interacted with year and included in a regression model predicting percentile net worth (married parents served as the reference category; demographic and income covariates were also included).The reference year was 2007.
Finally, we examine the mechanisms that may account for household type differences in net worth, including differences in assets, debts, and financial behaviors.In the interest of brevity, we focus our reported findings on differences in homeownership.We present a graph of homeownership rates by household type across income deciles.We estimated linear probability models of homeownership as a function of household type (with married parents serving as the reference category).We also re-estimated the quantile regression models of percentile net worth with homeownership included as a predictor, with the same set of demographic and income covariates as previous models.
All models except the unconditional quantile regression models were estimated using the scfcombo command in Stata.The scfcombo command was specifically designed for use with the publicly available SCF data sets, which included multiply imputed data with five implicates per household.(The command also adjusts for the sampling design of the SCF.)For the unconditional quantile regressions, we used the rifreg command in Stata with 200 bootstrap replications per implicate data set and then averaged the estimates together across the data sets.For predicted values, we use the margins command in Stata.

Results
A univariate analysis of median and percentile net worth indicates that wealth varies by parental status in complex ways (see Table 1; note that these estimates are from participants in the 2016 and 2019 SCF waves).Parents have almost three times as much median wealth as non-parents and were eight percentile points higher in the wealth distribution.Yet, when households are further classified by gender, marital status, and union type, the wealth advantage of parents over non-parents becomes obscured.The category of parent households includes both the household type with the highest level of median wealth (married parents, at $91,856) and that with the lowest level of median wealth (never-married mothers, at $241).Married childless couples are the second wealthiest category, and their median wealth ($57,771) is closer to that of married parents than any of the other non-parent household type categories.Single fathers have relatively high levels of wealth ($21,059), whereas cohabiting parents have low levels of wealth ($3,330).Single childless men have more wealth than unmarried cohabiting men, regardless of their parental status.These unadjusted estimates suggest that parenthood status per se is not associated with uniform disadvantages or advantages in wealth.
Results from regression analyses also indicate a complex association between parenthood and wealth.Panel A of Table 2 presents the results from regression models that included a simple dichotomous indicator of parental status, comparing childless adults with parents (the reference category).In the model that include only year effects (model 1), the difference between the two groups is 6.7 percentile points (at the mean, about $17,000).Most of this disparity can be accounted for by differences in observed demographic characteristics (model 2) and income (model 3).In the fully adjusted model (model 3), parenthood is associated with a 1.4 percentile point (p < 0.001) advantage in wealth.Panel B of Table 2 presents results by household type.The relative wealth advantage of parents over non-parents is less obvious in a more detailed comparison of household types.When only year fixed effects are included (model 1), marriedparent households have higher levels of wealth than all other household types (p < 0.001), including married childless couples.Consistent with expectations, accounting for demographic and income characteristics (models 2 and 3) greatly reduces, but does not eliminate, the wealth advantage enjoyed by married-parent households.Predicted percentiles of net worth by household type (results from model 3 of Table 2) are shown in Figure 1.After adjusting for demographic characteristics and income, married-parent households have higher levels of wealth than the other eight household types.Noticeably, though, the size of the wealth advantage of married-parent households over other household types did not show a clear pattern.For example, the wealth advantage of married-parent households over nevermarried mothers, married childless couples, and single men was about the same size.The categories with the largest wealth disadvantage vis-à-vis married-parent households were divorced mothers and cohabiting couples without children.Thus, once demographic characteristics and income covariates are included, parenthood was not uniformly associated with wealth advantages or disadvantages.
The wealth advantage associated with marriage did not characterize romantic unions more broadly.Cohabiting childless couples and cohabiting parents had similar levels of wealth; both had lower levels of wealth when compared with either married parents or married childless couples.
To summarize, on average, in models accounting for demographic characteristics and income, parents did not uniformly have higher levels of wealth, nor did households headed by a couple.Rather, the combination of marriage and parenthood resulted in the largest wealth advantage.

Differences across the Wealth Distribution
Turning to wealth differences across the wealth distribution (Table 3), results from the unconditional quantile regressions (with the full set of covariates for demographic characteristics and income) show that wealth disparities between parent and non-parent households (Panel A) were very small-less than one percentile point-at the bottom (25th percentile), middle (50th percentile), and top (75th percentile) of the wealth distribution.In contrast, the comparison of the full set of household types (Panel B) shows substantially more pronounced differences in wealth by household type and more variation in the magnitude of these differences across the wealth distribution.
In general, wealth disparities by household type were smaller at the top of the distribution, controlling for income and demographic characteristics.Among households near the 75th percentile of the wealth distribution, the estimated wealth rank of married-parent households was not different (at the p < 0.05 level of statissociological science | www.sociologicalscience.com tical significance) from that of single man households, and the difference between married parents and married childless couples was small.(Very few never-married mother households were near the 75th percentile of the wealth distribution, so we do not interpret the associated coefficient).Among the least wealthy households (those near the 25th percentile), single man households had lower levels of wealth than married-parent households.Notably, cohabiting parents, divorced mothers, single fathers, and cohabiting couples without children had lower levels of wealth than married-parent households across the wealth distribution.Moreover, the advantage of married-parent families over other parenting households was greatest near the middle of the distribution.
Among unpartnered adults, gender differences in wealth varied by parental status and place in the wealth distribution (Wald tests for differences by household type are available upon request).Near the 25th percentile, single fathers had a higher wealth rank than divorced mothers (p < 0.05), and single men had higher levels of wealth than single women (p < 0.05).Near the 75th percentile, gender differences among parents were less evident; single fathers and divorced mothers had similar levels of wealth.Gender differences among non-parent household heads remained, with single men having more wealth than single women near the 75th percentile.
In summary, the average wealth gap between parent and non-parent households did not vary across the wealth distribution, but this lack of aggregate differences obscures more nuanced patterns across household types.Differences between married-parent households and other household types were larger in the bottom and middle of the wealth distribution.Across the wealth distribution, single men without children had more wealth than single women without children, but the existence of a wealth gap between single fathers and divorced mothers varied across the wealth distribution.

Changes over Time: 1989 to 2019
Figure 2 shows unadjusted average percentiles of net worth by household type from 1989 to 2019.Surprisingly, wealth differences by household type were relatively stable over this period of changing wealth inequality and family demography.The figure shows that married-parent households consistently occupied the top position, whereas never-married mother households occupied the bottom of the hierarchy throughout the entire 30-year period.The stability of cohabiting parents and cohabiting childless couples in the wealth hierarchy is particularly interesting given the growth (and thus decrease in demographic selectivity) of this family type.
Because the demographic composition of household heads by household type changed between 1989 and 2019, we tested whether wealth gaps by household type remained stable after accounting for demographic characteristics and household income.Results from these models indicated a relative constancy of differences in wealth by household type.Among parent households, the interaction terms for unmarried parent and year were quite small (coefficients were close to zero), and none were statistically significant at p < 0.05.The wealth gap between married parents and unmarried childless adults also remained relatively static.

The Importance of Homeownership
To better understand the factors that account for differences in net worth by household type, we examined differences in assets, debts, and financial behaviors.We examined five mechanisms: attitude toward investment risk, savings habits, inheritance receipt, debt holdings, and asset ownership.
We found that these wealth-related mechanisms were predictive of net worth and varied by household type in expected ways.(Tables S3 and S4 of the online supplement contain selected results from these models.)For example, consistent with previous research (Conley and Ryvicker 2004;Hao 1996), we found that married couples were more likely than adults in other household types to report having received an inheritance and saving regularly.These mechanisms, however, did not account for the married-parent wealth advantage, and including them in models predicting net worth generally increased the size of the wealth difference between married parents and adults in other household types.We found that only one factor explained a substantial portion of the differences by household types in net worth: asset ownership, and more specifically homeownership.Thus, in the interest of brevity, we focus the discussion on our analyses of homeownership.
Homeownership rates by income decile and household type are shown in Figure 3.If the association between married parenthood and homeownership simply reflected the higher household incomes of married parents, then we would expect household type differences in homeownership to be muted within income deciles.Results do not support this conjecture, as homeownership rates are higher for married-parent couples than for other household types at every income decile.Moreover, the relative size of the homeownership gap between married-parent households and other household types is similar through most of the income distribution and smallest at the top of the distribution.
Linear probability models of homeownership confirm that married-parent households had much higher homeownership rates than all other household types (see Figure 4).Differences in homeownership by household type are only slightly attenuated between the unadjusted and adjusted model, reflecting that a relatively small share of the variation in homeownership is explained by household income and household head demographic characteristics.
We investigated whether household type gaps in homeownership had changed between 1989 and 2019 (see Table S5 of the online supplement).We found that these gaps were relatively stable.Only one of 10 interactions between unmarried childless household type and year was statistically significant at p < 0.05, and the coefficient was quite small.Four of the 10 interactions between unmarried parent and year were statistically significant at p < 0.05, but the coefficients on all but one of these interaction terms were also small.
Given these findings, we formally tested whether differences in homeownership can partially or fully account for parental status and household type differences in wealth.To do this, we included a dichotomous variable for homeownership in quantile regressions of percentile net worth. 3As shown in Table 4, we found that, after accounting for differences in homeownership, parent households have lower levels of wealth than non-parents at each point in the wealth distribution that we examined.Near the 25th and 50th percentiles, non-parents, relative to parents, were five percentile points higher in the wealth distribution (p < 0.001).Near the 75th percentile, non-parents are 1.4 percentile points higher in the distribution than parents (p < 0.001).These findings contrast with those from our previous models which showed a small average advantage of parents over non-parents (Table 2) and similarity in net worth between parent and non-parent households around the 25th, 50th, and 75th percentiles of the wealth distribution (Table 3).
Across household types, accounting for homeownership modifies the wealth differential in multiple ways (Panel B).First, accounting for homeownership greatly reduces or eliminates the size of the wealth disparity between married parents and other household types.The reduction in the relative wealth advantage of marriedparent households was largest near the 25th percentile but was also evident near the 50th and 75th percentiles.Second, across the wealth distribution, all four nonparent household types had wealth levels that were either equivalent to or greater than that of married parents after adjusting for homeownership.Near the 25th and 75th percentile, for example, childless married couples had wealth levels that were statistically indistinguishable from married parents' wealth; near the 50th percentile, married childless couples had slightly higher levels of wealth than married-parent households.Third, among the unmarried-parent categories, taking homeownership into account yielded a more heterogeneous pattern.Divorced mothers, even after accounting for homeownership, had lower levels of wealth relative to married parents at all three places in the wealth distribution (although, notably, the size of the disparity was substantially smaller than when homeownership was not modeled).In summary, our analyses indicate that homeownership was substantially more common among married-parent households than among any other household type, that these differences in homeownership were stable over time, and that differences in homeownership are particularly large in the middle and bottom of the U.S. wealth distribution.Moreover, the wealth advantage of married-parent households over other household types can largely be explained by differential rates of homeownership.At the mean, if childless married couples owned homes at the same rate as married parents, their predicted wealth would be greater than that of married-parent households.

Robustness Checks
We performed three main robustness checks, plus many additional checks.In the first, we transformed net worth using the inverse hyperbolic sine (IHS) function (Pence 2006) and re-estimated our models.Using the IHS transformation of wealth estimates the absolute wealth gap (measured in dollars) instead of the relative wealth gap (measured as position in the distribution).Findings were substantively the same as when we used percentiles of net worth.The only difference was that the IHS transformation showed no statistically significant difference in net worth between never-married mothers and married parents at the mean in models controlling for homeownership, whereas our preferred models indicate slightly higher levels of wealth for never-married mothers.
In the second robustness check, we defined net worth to include the value of vehicles and future pensions.Results were substantially the same and, if anything, indicated a slightly greater gap between single mothers and married parents, net of demographic and income characteristics.
In the third robustness check, we used a measure of wealth that excluded home equity.For households that do not own their homes, their net worth is the same regardless of the exclusion of home equity.For homeowner households, most have lower levels of wealth when home equity is excluded, but households with negative home equity have a higher net worth when home equity is excluded.Table S6 of the online supplement shows selected results from quantile regression models of this alternate measure of wealth.Panel A shows no difference in the revised net worth ranking between parents and non-parents at the 25th, 50th, or 75th percentiles, the same finding as from the models of total net worth (see Table 3).Panel B, with differences by household type, shows that married parents still had more wealth (excluding home equity) than cohabiting parents, divorced mothers, married non-parents, cohabiting non-parents, and single women at all three points in the distribution, but the gaps between these household types and married parents was smaller than in models of total net worth (including home equity).Single childless men ranked slightly higher in wealth at all three points of the distribution using this measure that excluded home equity.

Limitations
Our study has limitations.Our analyses are descriptive and use data from repeated cross-sectional surveys.Thus, we have an excellent portrait of how wealth holdings in the United States have changed over time and vary by household type across the population, but we cannot track how the wealth of any particular household has changed over time.The SCF data do not include information on wealth prior to the time of the survey, so we cannot examine how wealth trajectories are linked to household formation or parenthood.There may be other important mechanisms that account for household type differences that we are not able to investigate.For example, we do not have information on in vivo transfers from parents to adult children and how these may differ by household type.Insofar as married parents have unobserved characteristics that are associated with wealth, our results may overstate the wealth advantage associated with married parenthood.Additionally, geographic information from SCF is unavailable to researchers who are not employed by the Federal Reserve, so we cannot investigate geographic variation in wealth differences by household type.Moreover, we cannot ascertain how much of the variation in homeownership rates by household type is attributable to differences in norms around homeownership rather than to factors such as discrimination against single mothers and unmarried couples in housing and mortgage markets.Finally, there may be important differences by race and ethnicity in the association between household type and net worth or between household type and homeownership, but the sample sizes in the SCF data preclude investigation of such differences.

Discussion
Household net worth reflects the totality of households' expenditure, investment, and debt accrual decisions, and these decisions are shaped by the intersecting family-related identities of household heads.Our investigation of the wealth holdings of U.S. households from 1989 to 2019 addressed three elements heretofore overlooked in studies of parenthood and household wealth: the multiple identities inhabited by parents, variations across time and position in the wealth distribution, and wealth-building mechanisms.Although our analyses uncovered several interesting nuances in the associations between family characteristics and wealth, they also point to a succinct set of main findings: parents have a slight wealth advantage over non-parents that is driven by the large wealth advantage of married parents over other household types.The wealth advantage of married parents has been remarkably constant over the last 30 years and holds throughout the wealth distribution, although it is smaller at the top.Most notably, wealth differences by household type can be fully explained by differences in homeownership.
The intersection of parental and marital status resulted in a substantial wealth advantage, as married parents had higher levels of wealth than unmarried parents and adults without children.Accounting for demographic characteristics and income, the mean size of this wealth advantage ranged from approximately two percentile points higher than married childless couples and single men to nearly seven percentile points higher than cohabiting childless couples and divorced mothers.Wealth disparities between married parents and other household types were surprisingly stable over the 30-year period that we examined.Despite changing social selection into married parenthood, intensifying parental investments in children, and the growing prevalence of cohabitation, wealth disparities between married sociological science | www.sociologicalscience.com parents and other household types were stable over the 30-year period that we examined.
What accounts for this stable and sizable wealth advantage of married parents?It is not explained by differences in inheritance receipt, financial risk taking, savings habits, or debt holdings.Instead, we find that higher rates of homeownership fully account for the wealth advantage associated with married parenthood.
Married-parent couples were substantially more likely to own homes than adults in any other household type.Higher rates of homeownership were not driven by marriage or parenthood per se but by their combination; childless married couples and unmarried cohabiting parents had lower homeownership rates than married parents with the same income and demographic characteristics.Tellingly, the married-parent homeownership advantage was larger at the middle and lower end of the income distribution, indicating that higher homeownership rates for married parents do not just reflect their higher family incomes.
Our findings are consistent with Zelizer's (2000Zelizer's ( , 2013) ) foundational insights that intimate and relational identities, not just economic resources, shape economic transactions and consumptive behavior.Marriage and parenthood are two identities that together result in strong normative pressures and cultural expectations for homeownership (Mulder 2006(Mulder , 2013)).We postulate that homeownership achieves the primary identity work of married parenthood by displaying economic fitness for marriage and parenting, financial security and stability, and achievement of a middle-class lifestyle (Edin and Kefalas 2005;Dupuis and Thorns 2008).Given that homeownership is the mechanism by which most households accrue wealth in the United States (Wolff 2017), the norm of homeownership for married parents results in higher wealth levels for married parents than for adults in any other household type.
Wealth is an important determinant of adult and child well-being (Killewald et al. 2017).Our results suggest that the strong association between married parenthood and homeownership contributes to wealth inequalities.The concentration of married parenthood among highly educated adults (Gibson-Davis and Rackin 2014) and the wealth advantage of this family form raises new question for scholars concerned about wealth inequality and economic stratification.

Notes
1 Yamokoski and Keister's (2006) study is exemplary in its nuanced exploration of how gender, marital status, and parental status interact, but the study is restricted to the Baby Boomer cohort, which limits the generalizability of its findings, and it does not identify and examine cohabiting couples as a separate category.
2 Hao's (1996) research on households with children documents family structure differences in the composition of family wealth (the percentage of gross assets that are in savings, stocks, debts, homes, and other categories), but Hao's analysis does not test whether differences in wealth composition explain differences in wealth levels.Similarly, Grinstein-Weiss and colleagues (2008) describe household structure differences in the share of households with particular types of assets and the average values of those assets, but they do not incorporate this information further into their analysis.Yamokoski and sociological science | www.sociologicalscience.com (2006) include inheritance receipt as a predictor of net worth, but they do not describe whether inheritance receipt varies by gender, parental status, or family structure or show whether inheritance receipt widens or reduces family structure gaps.

Keister
3 Note that our regression model includes a dichotomous variable for homeownership, indicating that a home is included in the asset/debt portfolio.We do not include a continuous variable for home equity because home equity is a component of our outcome variable, net worth.Households with the same level of net worth can vary in their asset/debt portfolios, with some households owning their home and other households not holding this asset.If households had similar levels of total net worth regardless of their asset/debt profile composition, the homeownership variable would have a coefficient of 0, and inclusion of this variable would not explain any of the variation in net worth.

Figure 1 :
Figure 1: Predicted net worth percentiles, by household type.Notes: Error bars represent 95 percent confidence intervals.Estimates are adjusted for demographic and income variables and year fixed effects.N = 41, 625.

Figure 2 :
Figure 2: Mean percentile net worth by household type and year.

Figure 3 :
Figure 3: Homeownership by income decile and household type.

Figure 4 :
Figure 4: Predicted value of homeownership, by household type.Notes: Lines represent 95 percent confidence intervals.Unadjusted models control for year.Adjusted models control for demographic characteristics, income, and year.N = 41, 625.

Table 1 :
Median wealth and mean percentile net worth by household type, 2016 and 2019 Includes parents and non-parents.Values are in 2019 dollars.All estimates are weighted.Net worth adjusted for family size.

Table 2 :
Regressions of percentile net worth, by household type (N = 41, 625) Notes: All estimates are weighted.Statistical significance at p < 0.01 (two-tailed tests) is denoted as * .Standard errors are in parentheses.Demographic variables are household head age, age squared, race and ethnicity, and education.Income variables are family-size adjusted and include the linear and squared terms.Percentile net worth is based on size-adjusted family net worth, calculated within each year.Models in Panel B include all household types; coefficients for widowed individuals are suppressed to improve readability.

Table 3 :
Quantile regressions of percentile net worth, by household type (N = 41, 625) Notes: All estimates are weighted.Statistical significance at p < 0.01 (two-tailed tests) is denoted as * .Standard errors are in parentheses.All models include demographic and income variables, and year fixed effects.Models in Panel B include all household types; coefficients for widowed individuals are suppressed to improve readability.

Table 4 :
Quantile regressions of percentile net worth, by household type, adjusting for homeownership (N = 41, 625) All estimates are weighted.Statistical significance at p < 0.01 (two-tailed tests) is denoted as * .Standard errors are in parentheses.All models include demographic and income variables, homeownership, and year fixed effects.Models in Panel B include all household types; coefficients for widowed individuals are suppressed to improve readability. Notes: