Safety Nets and Scaffolds: Parental Support in the Transition to Adulthood

Abstract

Using longitudinal data from the Youth Development Study (analytic sample N = 712), we investigate how age, adult role acquisition and attainments, family resources, parent-child relationship quality, school attendance, and life events influence support received from parents in young adulthood. Parental assistance was found to be less forthcoming for those who had made greater progress on the road to adulthood, signified by socioeconomic attainment and union formation. The quality of mother-child and father-child relationships affected parental support in different ways, positively for mothers, negatively for fathers. School enrollment, negative life events, and employment problems were associated with a greater likelihood of receiving support. The findings suggest that parents act as “scaffolding” and “safety nets” to aid their children's successful transition to adulthood.

During the mid-twentieth century, most individuals could expect to take on adult roles and attain independence from their parents by their early-to-mid-20s. In recent decades the transition to adulthood has changed greatly. Structural and cultural changes have resulted in an extension of this life stage; young people spend more years pursuing education and take longer to establish careers and set up independent households in an economy that features unstable employment, low wages for the young, and high housing costs (Bell, Burtless, Gornick, & Smeeding, 2007; Settersten, Furstenberg, & Rumbaut, 2005). Some scholars now consider the contemporary transition to adulthood, rather than being concentrated in the early 20s, to span the ages of 18–34 (Furstenberg, Kennedy, McLoyd, Rumbaut, & Settersten, 2004).

Many parents are responding to the elongated transition by providing their young adult children with continued assistance (Aquilino, 2005; Fingerman, Miller, Birditt, & Zarit, 2009; Schoeni & Ross, 2005; Yelowitz, 2007). Some academics and public commentators have expressed concern that parental support in young adulthood prolongs dependence (e.g., Bahney 2006; Cote, 2000; Tyre, 2002). Others have argued that parental aid helps young people navigate the numerous challenges of early adulthood and supports their capacities for self-sufficiency (Aquilino, 2005; Eggebeen, 2005) by enabling young adults to develop the skills and experience needed to succeed in adult roles (Aquilino, 1999). In any case, for many families active parenting now extends beyond adolescence as the young adult's goal of independence becomes a “joint enterprise” (Scabini, Marta, & Lanz, 2006). But if the transition to adulthood is now a collaborative effort, under what conditions or situations do parents intervene and at what point do parents leave their children to manage on their own?

This study contributes to the literature by examining the conditions that lead to parental giving. By investigating factors that elicit support, we can better assess whether aid promotes progress in the transition to adulthood and responds to challenging circumstances, or whether it primarily reflects other factors, such as parental resources or relationship quality. This research is particularly important given the protracted transition to adulthood and the implications that this has for families as dependence upon parents extends into the 20s. Longitudinal survey data collected from young adults in the ongoing Youth Development Study (YDS) enable us to go beyond previous research by examining how age, as well as changes in circumstances and life events, affect parental giving over the course of this transition.

This study focused on housing and financial assistance as indicators of support. Although parental giving during this life phase also includes nonmaterial assistance, such as guidance, emotional nurturance, and practical help (Fingerman et al., 2009), we focused on these two forms of support because they signal material dependence or semidependence. Previous research has established that substantial numbers of young adults receive these forms of support from their parents during their 20s (Goldscheider & Goldscheider, 1999; Schoeni & Ross, 2005; Yelowitz, 2007). Coresidence may be an especially protective form of support in a difficult economy, as young adults who live with their parents are less vulnerable to poverty if they lose their jobs or suffer a wage decline (Bell et al., 2007). It is important to better understand how American families that try to help the next generation establish themselves as adults are responding to challenging conditions faced by young people.

We conceptualized and examined two circumstances surrounding support that may foster progress in the transition toward independent adulthood. We referred to the first as scaffolding, which we defined as temporary parental support that contributes to children's human capital or the achievement of specific socioeconomic attainment goals. We referred to the second as a safety net, which we defined as temporary, intermittent help given during difficult circumstances to minimize setbacks in the transition to adulthood.

Theoretical Background

Four theoretical perspectives informed our study of parental support. First, life course scholars have emphasized how normative expectations of age regulate people's behavior and shape perceptions of what they should or should not be doing at a particular age (Settersten, 1998). Patterns of parental intervention to help children negotiate the challenges of transitioning to adulthood reflect age norms of young adulthood (Settersten, 1998). For instance, financial and housing aid to adult children has been found to be most forthcoming when offspring are in their 20s (Silverstein, Bengtson, & Lawton, 1997). Likewise, reliance on parental help has been found to be less common and less favored as young people age. For example, coresidence has been shown to decline after age 25 (Schoeni & Ross, 2005), and over 90% of adults reported believing children should leave home by age 25 (Settersten, 1998).

Parental support, however, may have decreased with the adult child's age primarily because young people had taken on more adult roles signaling independence (Rossi & Rossi, 1990). Family formation, in particular, has been associated with greater independence from the natal family (Sarkisian & Gerstel, 2008; Smits, Van Gaalen, & Mulder, 2010). Still, some research found parental aid strongly age graded, even after accounting for life transitions, such as marriage, parenting, or student status, suggesting that age norms influenced giving decisions more than the accomplishment of adult statuses (Schoeni & Ross, 2005). Applying life course theory and research, we anticipated both age and progress in the transition to adulthood would influence the receipt of parental aid.

Second, we drew on the status attainment tradition, which has investigated the effects of familial resources and class origins on attainment (Kerckhoff, 1995; Sewell & Hauser, 1980). According to this perspective, those with more education and resources would be better able to invest in the development of their children's human, social, and cultural capital, contributing to a greater likelihood of socioeconomic success. Whereas this approach has focused primarily on adolescence, the prolonged transition to adulthood draws attention to the ways more advantaged parents may strengthen their children's prospects well after they leave high school, thus contributing to social class reproduction. Consistent with this perspective, previous research found that parents with more economic resources gave more financial help to their grown children (Hogan, Eggebeen, & Clogg, 1993; Schoeni & Ross, 2005). Families of higher SES were more likely to give money to meet living expenses, to purchase a house, or to obtain further education, all of which contributed to the greater human and financial capital of recipients (Semyonov & Lewin-Epstein, 2001; White, 1994). Beyond the availability of resources, Lareau (2003) found cultural differences in childrearing styles, with those from more class-advantaged backgrounds exerting concerted effort to cultivate children's achievement. These parenting styles may persist into young adulthood. Accordingly, we anticipated families with more resources would be more likely to supply their offspring with support as they invest in their children's attainment.

Third, existing theory and research have highlighted the importance of relational quality to intergenerational relations and giving. Intergenerational solidarity theory has predicted that family closeness would foster exchanges of support between generations (Silverstein et al., 1997). Adult children with closer relations with parents reported receiving more assistance from them (Eggebeen, 2005; Rossi & Rossi, 1990). Evidence has suggested that family relationships in adulthood have become “volitional,” with exchanges of support motivated by affection (Lye, 1996; Van Gaalen & Dykstra, 2006), perhaps because emotional solidarity and affection trigger intergenerational giving (Rossi & Rossi, 1990; Silverstein, Gans, & Yang, 2006). If a volitional model of parent-adult child support applies in this life stage, high quality relationships would increase the odds of young adults receiving parental support.

Other research has observed that in times of need, parents respond with support regardless of relationship quality (Ward & Spitze, 2007). This leads to our fourth theoretical frame. Contingency theory has posited family assistance to be more forthcoming when warranted by difficult life circumstances (Eggebeen & Davey, 1998; Fingerman et al., 2009; Silverstein et al., 2006). This approach conceptualized families as adaptive and responsive to members' needs and troubles. Parental help may be especially likely when offspring experience problems attaining or maintaining adult statuses (Ryff, Schmutte, & Lee, 1996). Parents may act as safety nets for their young adult children by offering aid that minimizes the negative effects of crises or difficulties, such as divorce, serious illness, or death of a spouse (Smits et al., 2010). The instability and rapid pace of change during early adulthood may make young adulthood a particularly vulnerable period necessitating a safety net more frequently than in other stages of life (Eggebeen, 2005).

Legitimate circumstances that garner parental aid may not be exclusively negative, but may also include positive opportunities that can bolster the adult child's potential for achievement and independence. Because the transition to adulthood has long-term consequences for socioeconomic attainment and security, parents may be especially inclined to offer help when it “scaffolds” offspring toward self-sufficiency through investment in human capital development (Lye, 1996). Indeed, parental attitudes toward continued support to adult children have been shown to depend on the circumstances, with greater approval for aid to those pursuing education (Aquilino, 2005; Goldscheider, Thornton, & Yang, 2001). Thus, support does not have to compete with an ideal of independence; it can be conceptualized as a step towards autonomy. We expected young adults would be more likely to receive parental support when they were students.

Variation in Parental Support by Race, Gender, and Family Structure

Because intergenerational coresidence varies by race in the United States, with Blacks, Hispanics, and Asians living with parents longer than Whites (Goldscheider & Goldscheider, 1999), we included race in our analysis. Gender was also included because research has shown gender differences in parental aid. For example, though sons and daughters were found to be equally likely to receive money from parents, sons were more likely to receive housing support (Ward & Spitze, 2007; White, 1994). Coresidence may be more comfortable for sons than daughters, as daughters perform more housework and are more closely supervised (Sassler, Ciambrone, & Benway, 2008).

We also controlled family size and structure. Family size has been found to influence the amount of intergenerational support; adult children with more siblings receive less money, as there are more people with whom to share parental resources (Conley, 2004). Previous research has also found continuously married parents to be more likely to provide support to grown children than single, divorced, or remarried parents (Aquilino, 2005). These variations have been attributed to differences in resources, relationship quality and contact, and ambiguous norms for stepfamilies (Amato, Rezac, & Booth, 1995; Aquilino, 2005; White, 1994).

Study Objectives and Hypotheses

This study contributes to scholarship on parental support during young adulthood in three ways. First, it examines factors that affect receipt of housing and financial support. Prior research has primarily focused on one or the other. Second, it is unique in its focus on provisions for basic living needs. Third, this longitudinal study moves beyond previous research in its ability to capture the effects of age, in addition to changes in life circumstances, on the receipt of parental support in young adulthood. The following hypotheses guided our analysis of parental support:

Hypothesis 1

Guided by life course theories, we anticipate that age norms of young adulthood and progress in attaining adult roles will affect parental giving. We expect that as respondents age, they will be less likely to receive financial and housing help from their parents. Moreover, young adults' progress in the transition to adulthood, as indicated by their family formation and socioeconomic attainment, will diminish the likelihood of receiving parental support.

Hypothesis 2

Drawing on the attainment perspective, we anticipate that the socioeconomic status of a young adult's family of origin will affect the likelihood of receiving parental support. We expect that respondents from families with higher income and those who have more highly educated parents will be more likely to receive assistance.

Hypothesis 3

Testing intergenerational solidarity theory and a volitional model of family support, we anticipate that respondents who report higher-quality parent-child relationships will be more likely to receive parental support.

Hypothesis 4

Guided by contingency theory, we anticipate that young adults' own circumstances and needs, as indicated by school attendance, employment status, and negative life events, will positively affect the receipt of support, as parents promote human capital acquisition (scaffolding) or respond to serious troubles (safety net).

Method

Data were drawn from the Youth Development Study (YDS), an ongoing longitudinal study beginning in 1988 (Mortimer, 2003). The initial sample (N = 1,010) consisted of randomly chosen ninth-grade students enrolled in the St. Paul public school district in Minnesota in the Fall of 1987. A probit analysis of the census tract characteristics of respondents (64% of invitees) and those who chose not to participate showed that girls and younger students were somewhat more likely to enroll. There was no significant selection in the original sample with regard to neighborhood characteristics, including racial composition, family structure, and socioeconomic status (Finch, Shanahan, Mortimer, & Ryu, 1991).

In the first four years of the study (1988–1991), participants completed questionnaires in their high schools. The data for 1992–2005 (Waves 5–16) were collected by mailed questionnaires asking about experiences in adolescence and the transition to adulthood, including family, work, school, and psychosocial characteristics. In 1988 and 1991, the respondents' parents were surveyed by mail to obtain accurate information regarding socioeconomic background.

We analyzed parental economic and housing support during a 9-year period, starting 6 years after high school completion (1997–2005). Respondents completed up to eight surveys at ages 24, 25, 26, 27, 29, 30, 31, and 32. We used data obtained from the parents and children from Waves 1–4 and 9–16. Wave 9 surveys took place in 1997, when most respondents were 24; Wave 16 data were collected in 2005, when they were 32 years old. In 2005, 710 respondents returned surveys. Women and Whites had higher retention than men and non-Whites; and youth without an employed parent during the year preceding the first wave (1987) had a higher risk of survey attrition than those who had at least one employed parent in 1987 (see Mortimer, 2003; Staff & Mortimer, 2007). In our analytic sample (N = 712), some 81% of participants identified as White and 19% identified as non-White (6.7% Black, 4.8% mixed race, 3.8% Hispanic, 1.7% Southeast Asian, .7% Other Asian, .7% Native American, and .7% other race/ethnicity).

Measures

Parental support

We examined the receipt of parental support during the latter period of transition to adulthood, after the traditional ages of college attendance. In each survey year (from age 24 in 1997 through age 32 in 2005), respondents were asked, “During the past year what share of your living expenses were covered by parents?” Economic support was a dichotomous variable indicating if any of the respondents' living expenses in the preceding year were covered by parents (0 = none, 1 = at least some). Housing support was also a dichotomous variable drawn from life-history calendars. We did not consider residing at home during a summer break (June–August), common for college students, as living with parents. Respondents who lived with their parents for at least one month, excluding summer months, in the 12 months preceding each survey were coded as receiving housing support (coded 1). We dichotomized the two parental support variables because in any single year most respondents received no support. Throughout the 8 waves, 52% of the respondents received economic support from parents at some point; 51% received housing support at some point. The data were converted to a person-year format, with variables indicating whether a respondent (i.e., child) received support in a given year (1) or not (0). Considering all person-years, 19% of the person-year units received economic support and 19% obtained housing support.

Table 1 lists the explanatory variables and their definitions. With the exception of age, time-varying explanatory variables refer to the year preceding measurement of the dependent variables. Background variables were measured in the first year of the study (1988).

Table 1.

Time-Varying and Background Variables: Descriptive Statistics (N=712)

Variables	Definition and categories
Time-varying variables: coded annually from wave 9 in 1997 through wave 16 in 2005
Age	Years following the first year of the time series (with age set at 0 at the starting point at age 24 in 1997)
Parenthood	Coded (1) if a parent; if not, coded (0)
Married	Coded (1) if living w/ spouse in any month during the past 12 months preceding the survey administration; if not, coded (0)
Cohabiting	Coded (1) if living w/ a partner in any month during the past 12 months preceding the survey administration; if not, coded (0)
R's highest level of education	Highest education completed: 1 = high school or less; 2 = technical/vocational school & associate degree; 3 = some college; 4 = bachelor's degree or above
R's income	Respondents' income during calendar year preceding the survey administration: 1 = under $5k; 2 = $5k–$9,999; 3 = $10k–$14,999; 4 = $15k–$19,999; 5 = $20k–$29,999; 6 = %30k–$39,999; 1 = $40K–$49,999; 8 = $50k–$59,999; 9 = $60k–$69,999; 10 = $70k–$79,999; 11 = $80k–$89,999; 12 = $90k–$99,999; 13 = $100k or more
School attendance	Coded (1) if attended school in any month during the past 12 months preceding the survey administration; if not, coded (0)
Negative events	Coded (1) if any of the following events happened during previous year: (a) breakup, (b) serious injury/illness, (c) spent time in jail, (d) assaulted, beaten up, robbed, or raped, (e) death of spouse/partner; if none, coded (0)
nemployment months	Coded as zero months unemployed (the reference category), 1–11 months unemployed, or 12 months unemployed during the past 12 months preceding the survey administration
Full-time work months	Coded as zero months full-time employment, 1–11 months full-time employment, or 12 months full-time employment (the reference category) during the past 12 months preceding the survey administration
Background variables		M	SD
Parental education in 1988	Parent's highest level of education: 1 = less than high school; 2 = high school; 3 = some college/junior college; 4 = 4-year college degree; 5 = some graduate school; 6 = master's degree; 7 = Ph.D. or professional degree	3.24	1.50
Family income in 1987	Family income: 1 = under $5k; 2 = $5k–$9,999; 3 = $10k–$14,999; 4 = $15k–$19,999; 5 = $20k–$29,999; 6 = $30k–$39,999; 1 = $40k–$49,999; 8 = $50k–$59,999; 9 = $60k–$69,999; 10 = $70k–$79,999; 11 = $80k–$89,999; 12 = $90k–$99,999; 13 = $100k or more	6.11	2.28
White	Respondent's race (1 = White, 0 = other)	.81
Male	Respondent's gender (1 = male, 0 = female)	.44
Stepparent family	Coded (1) if R is from a stepparent family; if not (0)	.15
Single parent family	Coded (1) if R is from a single parent family; if not (0)	.17
Other parent family	Coded (1) if R is from other types of family; if not (0)	.05
Continuously married parent family(reference category)	Coded (1) if R is from a continuously married parent family, if not (0)	.63
Number of siblings		1.43	1.16
Closeness to father during high school in 1988–1991	Average of the closeness to father index in waves 1 through 4 (each index ranges from 4–16)	9.49	2.71
Closeness to mother during the high school in 1988–1991	Average closeness to mother (computed in the same way as closeness to father)	11.54	2.49
Not a parent by 2005 (age 32)	Coded (1) if R is not a parent by 2005 (age 32); if not (0)	.36
Already a parent prior to 1997 (age 24)	Coded (1) if R is a parent prior to 1997 (age 24); if not (0)	.28
Became a parent (reference category) between 1997–2005	Coded (1) if R became a parent between 1997 (age 24) and 2005 (age 31); if not (0)	.36

Age and progress in the transition to adulthood

Because age and year were coterminous in this panel study, we measured age in years following the first year of the time series (with age set at 0 at the starting point at age 24 in 1997). The adult child's parenthood, relationship status (marriage, cohabitation, and neither), education level, and annual household income were time-varying (See Table 1). Respondents were considered parents if they had a child born in, or before, a given year. For most respondents, parenthood status did not vary during the period of observation; 64% of the respondents either had a child prior to age 24 (28%) or did not become a parent through the end of observation at age 32 (36%). To better represent parental timing in our models, two dummy background variables reflected the two groups whose parenthood status did not change over the period of observation (not a parent by 2005 (age 32) and already a parent prior to 1997 (age 24)); the reference category consists of those who became parents during this period). Respondents were considered married if married at least one month of the year; and cohabiting if they cohabited for at least one month and were not married during the prior year (neither cohabiting nor married is the reference).

Time-varying life circumstances associated with parental support

Life circumstances, including school attendance, negative life events, and employment states, were measured as time-varying dichotomous variables (See Table 1). Respondents were considered to be in school (coded 1) if they reported attending school for at least one month of the year. Respondents were considered to have experienced a negative event (coded 1) if one or more of several events, shown in Table 1, happened during the previous year. To measure employment problems, we examined periods of unemployment (1–11 months unemployed, and 12 months unemployed; zero months unemployed is the reference category), and periods without full-time work (zero full-time work during the year, and 1–11 full-time work months; the reference category is 12 full-time work months).

Background variables

With the exception of parental timing, described above, background characteristics were derived from the Wave 1 child and parent surveys in 1988 (See Table 1). Though most of these could change over time (e.g., structure of the family of origin), for the purpose of this analysis they were considered background characteristics, referencing states in adolescence at the beginning of the study. The higher value of the mother's or father's education indicated the family's educational background. Family income (in 1987) was represented by a 13-point scale. Race, gender, stepparent family, single parent family, and other (the continuously married parent family is the reference) were dichotomous; number of siblings was continuous.

Parent-adolescent relationship

During high school, respondents were asked about their feelings of closeness to their fathers and mothers. An additive index, based on the following items, was constructed for each year: (a) “How close do you feel to him/her?” (1 = Not close at all to 4 = Extremely close), (b) “When you are faced with personal concerns and decisions, do you talk them over with him/her?” (1 = Never to 4 = Often), (c) “How often does he/she talk over important decisions that he/she has to make with you?” (1=Never to 4 = Often), and (d) “How often does he/she listen to your side of an argument?” (1 = Never to 4 = Often). Indices from the four years were averaged to form two closeness indices, one for father (α = .86) and one for mother (α = .86).

Analytic Strategy

We used multilevel models to estimate the effects of individual background characteristics, age, time-varying life circumstances, and indicators of progress in the transition to adulthood on parental economic and housing support. In our analyses, time points were nested within individuals. The assumption made for statistical inference that error terms are independent was not likely to be satisfied, because error terms associated with the same person are likely to be positively correlated (Raudenbush & Bryk, 2002). A multilevel framework appropriately addressed the nested nature of the data. Two-level hierarchical generalized linear models (HGLM) were estimated with a Bernoulli distribution log-link function; the population average models with robust standard errors were reported (Raudenbush & Bryk, 2002).

This multilevel analysis had several advantages over conventional analyses such as ordinary least squares (OLS) regression (Raudenbush & Bryk, 2002). It took into account not only the correlated residuals, but by mean-centering the time-varying explanatory variables, controlled for a host of potential sources of spuriousness. Another advantage of the multilevel framework was that it handled unbalanced and incomplete data appropriately in estimating the parameters. The estimation procedure essentially weighted for the number of observations per person; those respondents who participated in fewer than eight waves contributed less to the overall estimates.

The multilevel framework incorporated covariates that vary with respect to time in the Level 1 model. Covariates considered invariant with respect to time (background variables) were incorporated into Level 2 of the model as predictors of person-specific intercepts and age slopes.

A separate set of HGLMs was estimated for each dependent variable, parental economic support, and parental housing support. The first step in estimating this model was to create a person-year data set for each survey year from 1997 through 2005; there were 4,387 observations for the economic support models and 4,414 observations for the housing support models. The person-year was the unit of analysis at Level 1 and each person-year was nested within an individual at Level 2 (N = 712 respondents). Sample loss is due to attrition (those who dropped out of the study by 1997 would have no Level 1 variables during the period of observation), as well as the absence of particular background variables.

The values of the dependent variable were transformed from 1 (receiving parental support) and 0 (not receiving parental support) to the linearized log odds of receiving parental support. Because the outcome was dichotomous, we specified a Bernoulli model with a logit link. The Level 1 variance was fixed by the proportion receiving parental support. Consequently, the variance could not be apportioned within and between individuals as is customary with hierarchical linear models (HLM). HGLMs also differ from standard HLM models because an appropriate deviance statistic cannot be estimated to assess change in model fit. The Level 1 equation of the final full model for receiving parental support for person i at time t was written:

$$Y_{\mathit{it}}=\text{log}[P_{\mathit{it}}∕(1-P_{\mathit{it}})]={\beta }_{0i}+{{\beta }_{1i}}^{\ast }\left({\text{time}}_{\mathit{it}}\right)+{{\beta }_{2i}}^{\ast }\left({\text{parenthood}}_{\mathit{it}}\right)+{{\beta }_{3i}}^{\ast }\left({\text{married}}_{\mathit{it}}\right)+{{\beta }_{4i}}^{\ast }\left({\text{cohabiting}}_{\mathit{it}}\right)+{{\beta }_{5i}}^{\ast }\left({\text{highest level of education}}_{\mathit{it}}\right)+{{\beta }_{6i}}^{\ast }\left({\text{income}}_{\mathit{it}}\right)+{{\beta }_{7i}}^{\ast }\left({\text{school attendance}}_{\mathit{it}}\right)+{{\beta }_{8i}}^{\ast }\left({\text{negative events}}_{\mathit{it}}\right)+{{\beta }_{9i}}^{\ast }(1–11{\text{months unemployed}}_{\mathit{it}})+{{\beta }_{10i}}^{\ast }\left(12{\text{months unemployed}}_{\mathit{it}}\right)+{{\beta }_{11i}}^{\ast }\left({\text{zero month full-time work}}_{\mathit{it}}\right)+{{\beta }_{12i}}^{\ast }(1–11{\text{full-time work months}}_{\mathit{it}})$$

Both Y, referring to the odds of receiving parental support, and the β parameters were uniquely estimated for each person i. β_1i is the average rate of change in the odds of receiving parental support per year, and time was measured as years passed since 1997. All Level 1 predictors were centered around their means (person mean over the 8-year period) in order to control for between-individual differences in time-varying variables and thus minimize sources of unobserved heterogeneity that might influence the odds of receiving parental support (Halaby, 2003; Raudenbush & Bryk, 2002).

The Level 2 equations of the final full model were written as:

$${\beta }_{0i}={\gamma }_{00}+{{\gamma }_{01}}^{\ast }\left({\text{parental education}}_{\mathit{i}}\right)+{{\gamma }_{02}}^{\ast }\left({\text{family income}}_{\mathit{i}}\right)+{{\gamma }_{03}}^{\ast }\left({\text{White}}_{\mathit{i}}\right)+{{\gamma }_{04}}^{\ast }\left({\text{male}}_{\mathit{i}}\right)+{{\gamma }_{05}}^{\ast }\left({\text{stepparent}}_{\mathit{i}}\right)+{{\gamma }_{06}}^{\ast }\left({\text{single parent}}_{\mathit{i}}\right)+{{\gamma }_{07}}^{\ast }\left({\text{other parent}}_{\mathit{i}}\right)+{{\gamma }_{08}}^{\ast }\left({\text{sibling}}_{\mathit{i}}\right)+{{\gamma }_{09}}^{\ast }\left({\text{closeness to father}}_{\mathit{i}}\right)+{{\gamma }_{010}}^{\ast }\left({\text{closeness to mother}}_{\mathit{i}}\right)+{{\gamma }_{011}}^{\ast }\left({\text{not a parent by 2005}}_{\mathit{i}}\right)+{{\gamma }_{012}}^{\ast }\left({\text{already a parent prior to 1997}}_{\mathit{i}}\right)+{\mu }_{0i}$$

$${\beta }_{1i}={\gamma }_{10}+{{\gamma }_{11}}^{\ast }\left({\text{parental education}}_{\mathit{i}}\right)+{{\gamma }_{12}}^{\ast }\left({\text{family income}}_{\mathit{i}}\right)+{{\gamma }_{13}}^{\ast }\left({\text{White}}_{\mathit{i}}\right)+{{\gamma }_{14}}^{\ast }\left({\text{male}}_{\mathit{i}}\right)+{{\gamma }_{15}}^{\ast }\left({\text{stepparent}}_{\mathit{i}}\right)+{{\gamma }_{16}}^{\ast }\left({\text{single parent}}_{\mathit{i}}\right)+{{\gamma }_{17}}^{\ast }\left({\text{other parent}}_{\mathit{i}}\right)+{{\gamma }_{18}}^{\ast }\left({\text{sibling}}_{\mathit{i}}\right)+{{\gamma }_{19}}^{\ast }\left({\text{closeness to father}}_{\mathit{i}}\right)+{{\gamma }_{110}}^{\ast }\left({\text{closeness to mother}}_{\mathit{i}}\right)+{\mu }_{1i}$$

$${\beta }_{\mathrm{x}i}={\gamma }_{\mathrm{x}0}$$

β_0i, the individual intercept at t = 0 (i.e., at age 24), and β_1i, the rate of change over time, were modeled as a function of specified individual characteristics (γ). In order to better represent the two large groups that did not change their parenthood status during the observation period, we included two dummies (not a parent by 2005 (age 32) and already a parent prior to 1997 (age 24)) to predict the Level 1 intercept. Both groups had a score of zero in all time-varying, mean-centered parenthood spells although their likelihood of receiving support was quite different. Due to multicollinearity (overlap of “parenting during the observation period,” the reference category of the dummies, and the Level 1 parenting variable), we did not include the dummies to predict the age slope. The random error components, μ_0i and μ_1i, indicated the individual variation around the intercept and the average rate of change.

We did not model other Level 1 parameters (β_2i through β_12i) as a function of specific individual characteristics due to the absence of theory as to why these parameters should be considered related to those characteristics and because of the limited degrees of freedom. Thus, the effects of other time-varying covariates, β_xi, were fixed across people and did not have an error term specified at Level 2.

To test our hypotheses, variables were entered in a staged sequence, guided by our conceptual model (Raudenbush & Bryk, 2002). In Model 1 both the Level 1 intercept and the age slope were set to vary randomly, but no attempt was made to predict this variation (Raudenbush & Bryk, 2002, p. 26). In Model 2, we considered progress in the transition to adulthood; the receipt of support was modeled as a function of age, parenthood, relationship status, and socioeconomic attainments at Level 1. Family socioeconomic characteristics and structure, race, gender, number of siblings, parent-adolescent relationship quality, and the dummy variables indicating invariant parental status (always and never during the period of observation) also entered into Model 2 at level 2. In Model 3, to test the contingency model, and the scaffolding and safety-net hypotheses, we added the youth's life circumstances, including educational attendance, negative events, and employment states at Level 1.

Findings

Tables 2 and 3 present the population-average model results, which indicate how a change in each predictor affects the overall population odds of receiving any parental economic support and housing support, respectively, from age 24 (in 1997) through age 32 (in 2005).

Table 2.

HGLM Results for Receiving Parental Economic Support (Population Average Models)

	Model 1			Model 2			Model 3
	β	SE	Odds	β	SE	Odds	β	SE	Odds
Intercept	−1.40	.05	.25***	−2.39	.36	.09***	−2.42	.36	.09***
Age	−.16	.01	.85***	.03	.07	1.03	.02	.07	1.02
Parenthood				.20	.11	1.22	.18	.11	1.20
Married				−.69	.12	.50***	−.64	.12	.53***
Cohabiting				−.46	.09	.63***	−.43	.10	.65***
R's highest level of education				−.16	.08	.85*	−.12	.07	.88
R's income				−.08	.02	.92***	−.05	.02	.95**
School attendance							.42	.08	1.52***
Negative event							.15	.08	1.16
1–11 months unemployed							.27	.11	1.31*
12 months unemployed							.37	.28	1.44
Zero full-time work months							.29	.11	1.34**
1–11 full-time work months							.25	.09	1.29**
Level 2 variables predicting the intercept
Parental education				.12	.04	1.13**	.12	.04	1.13**
Family income				.02	.03	1.02	.02	.03	1.02
White				−.07	.13	.93	−.07	.13	.93
Male				.11	.10	1.12	.12	.10	1.12
Stepparent				−.45	.13	.64***	−.45	.13	.64***
Single parent				−.13	.15	.88	−.13	.15	.88
Other parent				.19	.24	1.21	.20	.24	1.22
Number of siblings				−.07	.04	.93	−.07	.04	.93
Closeness to father				−.03	.02	.97	−.03	.02	.97
Closeness to mother				.06	.02	1.06**	.06	.02	1.06**
Not a parent by 2005				.72	.12	2.05***	.71	.12	2.03***
Already a parent prior to 1997				−.33	.13	.72*	−.34	.13	.71*
Level 2 variables predicting the slope of age
Parental education				−.01	.01	.99	−.01	.01	.99
Family income				.01	.01	1.01	.01	.01	1.01
White				−.02	.03	.98	−.03	.03	.97
Male				.02	.02	1.02	.03	.02	1.03
Stepparent				.02	.03	1.02	.01	.03	1.01
Single parent				.10	.03	1.11**	.09	.03	1.10**
Other parent				.05	.05	1.05	.04	.05	1.04
Number of siblings				−.01	.01	.99	−.01	.01	.99
Closeness to father				.00	.00	1.00	.00	.00	1.00
Closeness to mother				−.01	.00	99***	−.01	.00	99***

Note: N = 4387 (Level 1), 712 (Level 2)

^*p < .05.

^**p < .01.

^***p < .001.

Table 3.

HGLM Results for Parental Housing Support (Population Average Models)

	Model 1			Model 2			Model 3
	β	SE	Odds	β	SE	Odds	β	SE	Odds
Intercept	−1.37	.05	.25***	−1.05	.33	.35**	−1.07	.33	.34**
Age	−.20	.01	82***	.01	.08	1.01	.00	.08	1.00
Parenthood				.16	.12	1.17	.14	.12	1.15
Married				−1.12	.13	33***	−1.06	.13	35***
Cohabiting				−.81	.11	44***	−.78	.11	.46***
R's highest level of education				−.14	.07	.87*	−.11	.07	.90
R's income				−.00	.02	1.00	.01	.02	1.02
School attendance							.31	.08	1.36***
Negative event							.22	.08	1.25*
1–11 months unemployed							.15	.11	1.17
12 months unemployed							−.02	.29	.98
Zero full-time work months							.13	.11	1.14
1–11 full-time work months							.14	.09	1.15
Level 2 variables predicting the intercept
Parental education				−.14	.03	.87***	−.14	.03	.87***
Family income				−.03	.03	.97	−.03	.03	.97
White				−.05	.12	.95	−.05	.12	.95
Male				.19	.10	1.21*	.19	.10	1.21*
Stepparent				−.24	.13	.79	−.24	.13	.78
Single parent				−.28	.15	.75	−.29	.15	.75
Other parent				.09	.21	1.09	.09	.21	1.10
Number of siblings				−.01	.04	.99	−.01	.04	.99
Closeness to father				−.04	.02	.96*	−.04	.02	.96*
Closeness to mother				.04	.02	1.04*	.04	.02	1.04*
Not a parent by 2005				.91	.12	2.49***	.91	.12	2.48***
Already a parent prior to 1997				−.48	.13	.62***	−.47	.13	.62***
Level 2 variables predicting the slope of age
Parental education				.01	.01	1.01	.01	.01	1.01
Family income				.00	.01	1.00	.00	.01	1.00
White				−.05	.03	.95	−.05	.03	.95
Male				−.01	.03	.99	.00	.03	1.00
Stepparent				.04	.04	1.04	.04	.04	1.04
Single parent				.05	.04	1.05	.04	.04	1.04
Other parent				−.05	.06	.95	.05	.07	.95
Number of siblings				−.01	.01	.99	−.01	.01	.99
Closeness to father				.00	.00	1.00	.00	.00	1.00
Closeness to mother				−.01	.00	.99*	−.01	.00	.99*

Note. N = 4414 (Level 1), 712 (Level 2)

^*p < .05.

^**p < .01.

^***p < .001.

Economic Support

Turning first to economic support (Table 2), Model 1 confirmed Hypothesis 1, that economic support decreases as respondents age. On average, the odds of receiving economic support decreased by 15% per year (p < .001). Model 2 also supported Hypothesis 1 in that some indicators of progress in the transition to adulthood lessened the likelihood of receiving economic support from parents. Being married (versus neither married nor cohabiting) decreased the odds of receiving economic support by 50% (p < .001), and cohabitation decreased the odds by 37% (p < .001). Moreover, each increase in the young adult's educational attainment and household income (across the 13 categories) decreased the odds of support by 15% (p < .05) and 8% (p < .001), respectively. These findings were generally robust across Models 2 and 3 and confirmed our hypotheses that age and progress in the transition to adulthood, as indicated by union formation and socioeconomic attainment, diminish economic support from parents. The respondent's level of education had a significant negative independent effect on the odds of receiving economic support in Model 2 but the effect diminished when the young adult's changing life circumstances were controlled in Model 3.

Although parenthood is an important marker of transition to adulthood, having a child between ages 24 and 32 did not affect the odds of receiving economic support. Effects of the dummy variables representing those whose parental status was time invariant (those who already had a child before age 24, and those who remained childless through age 32), however, suggested that parental status matters for support. Compared to those who became a parent between ages 24–32, those who remained childless throughout the study period had much higher odds of receiving economic support at age 24 (by 105%, p < .001) and those who already had a child before age 24 had lower odds of receiving economic support at age 24 (by 28%, p < .05).

Model 2 also provided partial support for Hypothesis 2 that family socioeconomic background affects the likelihood of receiving parental support. Parental education was found to be positively related to support at the start of the observation period. Each increase in parental educational level increased the odds of support by 13% at age 24 (p < .01). Family income did not significantly influence financial support.

Hypothesis 3 focused on the quality of the parent-adult child relationship as a correlate of support. Although closeness to father was not associated with economic support at age 24, closeness to mother increased the odds of support. Each increase in closeness to mother increased the odds by 6% (p < .01). The positive effect of closeness to mother on the intercept was stable across Models 2 and 3. Respondents who were closer to their mothers, however, had significantly steeper declines in economic support as they aged.

Model 3 showed partial support for Hypothesis 4 that parental support is contingent on the young adult's changing life circumstances. Variables introduced at Level 1, indicating needs for scaffolding and safety nets, increased parental support. School attendance increased the odds of receiving economic support by 52% (p < .001), supporting our scaffolding hypothesis. “Legitimate need,” as indicated by employment problems, was associated with greater likelihood of economic support. Partial-year unemployment was associated with significantly higher odds of receiving economic support. Those who were partly unemployed during the preceding year were 31% more likely to receive economic support than those who were never unemployed (p < .05). Full-year unemployment, however, did not change the odds. No full-time work or only partial-year full-time work also increased the odds of receiving support (versus full-year full-time work) by 34% (p < .01) and 29% (p < .01), respectively. Contrary to the contingency hypothesis, experiencing one or more negative events during the preceding year did not change the odds of receiving parental financial support.

Whereas children from single parent families were not less likely to receive economic support than those from continuously married parent families at the start of observation, they experienced less decline in support with age than did children from continuously married families. Those from stepfamilies appeared to be at a disadvantage at the outset. The odds of receiving economic support at age 24 for those from a stepparent household were 36% lower than respondents from a continuously married parent household (p < .001). Neither gender nor number of siblings were significant factors in receiving economic support.

Housing Support

Table 3 displays the results for housing assistance, or more specifically, coresidence in the parental home. There was substantial similarity in findings for the two sets of models.

Hypothesis 1 was supported, as age had a strong negative effect on housing support (Model 1, p < .001). Like economic support, in Model 2, union formation, but not the onset of parenthood between ages 24 and 32, significantly decreased the odds of coresidence. Being married (versus neither married nor cohabiting) decreased the odds of coresidence by 67% (p < .001). Cohabiting (versus neither married nor cohabiting) decreased the odds of coresidence by 56% (p < .001). Also similar to the results for economic support, those who became a parent between ages 24–32 had significantly lower odds of receiving housing support at the beginning of the observation period than those who remained without a child by age 32, but much higher odds than the early parents, those who already had a child before age 24. Progress in the transition to adulthood, as indicated by change in respondents' educational attainment and income, were not significantly associated with the odds of coresidence.

Contrary to Hypothesis 2, that greater parental resources would increase the likelihood of housing support, young adults whose parents were more highly educated actually received less residential support than those whose parents had less education. Family income did not significantly predict the odds of receiving housing support.

With respect to Hypothesis 3, implicating the quality of parent-child relationships, we found the opposite impacts of closeness to father and mother on the odds of receiving housing support; closeness to father decreased the odds and closeness to mother increased the odds. Like economic support, closeness to mothers in adolescence predicted a steeper decline in residential support with age.

Relevant to our fourth contingency hypothesis, Model 3 showed that school attendance increased the odds of living with parents by 36% (p < .001). Unlike economic support, negative events increased the odds of coresidence by 25% (p < .01). Changes in unemployment or full-time-work problems were not significantly associated with the odds of coresidence.

With respect to the variables entered as controls, Model 2 showed that men were 1.2 times more likely to receive housing support than women (p < .05) at the outset; men's greater likelihood of living with parents was apparent in both Model 2 and Model 3. Variables representing structure of the family of origin did not predict the odds of receiving housing support at age 24 or the rate of change in residential support with age. Number of siblings had no significant effects on housing support.

Discussion

The extended transition into adulthood has implications for families as many parents continue to support their young adult children. Although almost half of young adults in the YDS received at least one of the two forms of parental assistance—financial or housing support— in their early 20s, only 10–15% received help in their early 30s. The robust effect of age suggests a strong normative framework supporting the independence of children as they grow older.

Also consistent with the life course perspective, the child's own progress toward becoming an adult, as indicated by intimate partnership and income attainment, diminished parental aid. Parents were especially disinclined to provide money or housing to young adult children who were married or cohabited. This finding is consistent with other research that has found lower levels of intergenerational support from parents to married adult children compared to their single counterparts (Hogan et al., 1993; Sarkisian & Gerstel, 2008). Those who remained childless through the study period received more support than those who had children. These results suggested that family formation was viewed as an important marker of adulthood, signaling that independence and autonomy were expected. Our data cannot assess who initiated support. Possibly, parents viewed offspring who had begun to form families as having transitioned into adulthood, and believed it was no longer appropriate to provide for them. Or perhaps these young adults felt it was inappropriate to seek or accept parental help once they had formed their own families. Additionally, married or cohabiting respondents may have had less need for parental aid because they had partners to turn to when problems arose.

From the status-attainment perspective, we anticipated that young adults from families with greater resources would be more likely to receive financial assistance from their parents. Yet we found parents across income groups were similarly inclined to provide at least some money for living expenses to their adult children. We found, however, more highly educated parents to be more likely to provide money. This pattern offered some evidence of a social class difference in cultural expectations for prolonged support during young adulthood and the possibility that more highly educated parents viewed their young adult child's success as a kind of joint project which required their “concerted cultivation” (Lareau, 2003) and continued investment. The large differences by social class reported in other studies (Hogan et al., 1993; Schoeni & Ross, 2005) likely reflect the amount of money given, or gifts for extras or education, which our data did not capture.

We found mixed evidence for a volitional model of parent-adult child relationships that would encourage assistance depending on prior relationship quality; close relationships with mothers fostered both kinds of support initially, but relationship quality with fathers had either no effect (economic support) or a negative effect (housing support). Perhaps close relationships with mothers facilitated young adults asking for help, or mothers who were closer to their children were more aware of their needs. Although those who were closer to their mothers in adolescence received more help in early adulthood, they had a steeper decline in support as they aged. It may be that in addition to this early material assistance, those who were closer to their mothers also received other types of parental aid, such as advice, emotional support, and practical help that could have contributed to their ability to become self-sufficient.

Effects for relationships with fathers ran counter to solidarity or volitional model predictions. One plausible interpretation for the unanticipated negative effect of closeness to father on housing support is that movement towards early independence fostered good relationships, or alternatively, that strong relationships with fathers in adolescence encouraged earlier residential independence. These findings, taken together, suggest that relationship quality with mothers and fathers may be based on different criteria, with distinct implications for intergenerational support. Further research in this area is needed. We may have found more consistent evidence that solidarity activates aid if we had examined other forms of help, such as services, occasional gifts, or emotional support, in addition to housing or supplements for living expenses.

Our findings regarding safety nets and scaffolding are particularly revealing. Lending credence to our hypotheses, negative life events and school attendance instigated parental assistance. The situational nature of parental assistance supported the contingency theoretical approach and mirrored findings of other studies of intergenerational help (Fingerman et al., 2009; Silverstein et al., 2006; Ward & Spitze, 2007). Different types of problems, however, elicited distinct forms of help. Financial assistance was responsive to situations representing temporary unemployment, lasting less than a full year, suggesting that when young adults are temporarily derailed by interruptions in employment, parents come to the rescue with money for living expenses. Yet parents did not increase financial support to those adult children who experienced more serious, chronic unemployment. Thus, our data suggested that parents acted as safety nets that helped their children get back on their feet, rather than offering long-term support that could prolong dependence as proposed in the media of late (e.g., Bahney 2006; Tyre 2002). Parents, though, were more likely to support adult children financially who did not work full-time, whether temporarily or not. This may indicate parents' willingness to partially provide for their young adult children as they encounter difficulties securing full-time jobs or as they take longer to explore work options.

Housing support was forthcoming for those who experienced other problems and crises not related to employment, such as divorce, serious illness, or death of a spouse. Financial support may be a more easily transferable asset that is responsive to temporary problems like intermittent unemployment. In contrast, moving in with parents could be a more consequential response in reaction to a serious negative life event, such as divorce or separation (signaling more durable shifts in resources and statuses) or serious illness that may require multiple forms of help. Whether in the form of financial or housing assistance, results from this study demonstrated that parents served as safety nets when their young adult children encountered difficult circumstances.

The significant positive effects of school attendance across models supports the scaffolding hypothesis. Student status promoted both financial and housing help, independent of age, family resources, parent-child relationship quality, and important life circumstances. Parents helped their children to take advantage of opportunities to build human capital and improve their chances for achievement by providing for them while they acquired more education.

Consistent with previous research, men were more likely than women to receive residential help from parents. Previous research has suggested that gendered expectations have contributed to coresident daughters having more household responsibilities and being more closely monitored and controlled than sons (Sassler et al., 2008). If women have greater desire to escape onerous conditions in the parental home, this could motivate earlier family formation and other activities that reduce the likelihood of coresidence.

Family structure had some bearing on receipt of parental aid. Those from stepfamilies were less likely to receive financial transfers than those from continuously married parent families. Others have explained this pattern as deriving from the diffusion of responsibility among custodial and noncustodial parents, as well as less close parent-child relations (Aquilino, 2005).

Conclusion

A central finding of this study was that parental support in young adulthood depended on contemporaneous circumstances. Young adults' life events and needs influenced parents' willingness to help them: when young people were moving toward socioeconomic attainment or when they have experienced a crisis, they were more likely to receive financial and housing assistance. In this way, parents served as collaborators in their children's transition to adulthood, acting as scaffolding systems to help young people reach their goals and as safety nets to catch them before they fell too far.

This study provides evidence that parental assistance in young adulthood promoted progress toward self-sufficiency under changed structural conditions. In an economy that requires advanced education for good jobs, parents were more likely to aid their children when they were students. As the labor market offers fewer opportunities for stable, full-time, well-paid work for the young, parents often filled in when needed. As young people progressed towards becoming adults, parents stepped back.

A few limitations of this study warrant attention. Respondents in this study came from St. Paul public schools, and were recruited during a time when there was less racial and ethnic diversity than today. Regional samples have been used in other important studies of parental giving (Fingerman et al., 2009; Rossi & Rossi, 1990), yet future research would benefit from nationally representative and more diverse samples. This study probably underrepresents the wealthiest families, many of whom send their children to private schools. Also, time-varying measures of relationship closeness may have yielded more support for the volitional hypothesis.

Still, these findings validate the assertion that intergenerational relations may become increasingly important as people delay marriage and as marriage becomes less stable (Bengtson, 2001). Perhaps individuals rely on their partners when they have them, but turn to their parents when they do not have partners. Alternatively, parents may withdraw support to uphold the boundaries of the newly forming nuclear family, viewing self-reliance as essential for nuclear family well-being. In either case, even when young people delay or forego partnering, or experience relationship instability, many have family to turn to in times of need. Indeed, the findings suggest that families are adaptive to changing situations that threaten the well-being of their members.

Prolonged family support is not without its drawbacks. It may play a role in solidifying, or even exacerbating, economic inequalities and may cause strain for families with few resources. In addition, these findings hold implications for policy. Young adults without access to families who can provide for them during difficult times or when they need further education may find themselves at a decided disadvantage. Additionally, policy makers and institutions should keep in mind that dependency does not necessarily end at the age of majority, but that parents may extend their support to children a decade or more beyond this. Thus, families are absorbing some of the problems associated with an economy that requires more education but offers less stable employment and lower wages for young people. In any case, what is apparent is the ongoing significance of family, especially with respect to the increasing importance and relevance of intergenerational assistance.