Three Generation Family Households: Differences by Family Structure at Birth

Abstract

Using data from the Fragile Families and Child Well-being Study (N=4,898), this study investigates how the share, correlates, transition patterns, and duration of three generation households vary by mother’s relationship status at birth. Nine percent of married mothers, 17 % of cohabiting, and 45% of single mothers live in a three generation family household at the birth of the child. Incidence over time is much higher and most common among single mother households, 60% live in a three generation family household in at least one wave. Economic need, culture, and generational needs are associated with living in a three generation household and correlates vary by mother’s relationship status. Three generation family households are short lived and transitions are frequent. Kin support through coresidence is an important source of support for families with young children and in particular families that are unwed at the birth of their child.

Vern Bengtson (2001) in his Burgess Award Lecture to the National Council on Family Relations boldly stated: “For many Americans multigenerational bonds are becoming more important than nuclear family ties for well-being and support over the course of their lives” (Bengtson, 2001: 5). Whether more important than nuclear family ties, multigenerational ties are an essential part of the family system (Swartz, 2009) and multigenerational family households (where two or more adult generations coreside) have recently increased in prevalence (Taylor et al, 2010). This paper focuses on one type of type of multigenerational household, a three generation family household where a grandparent, parent, and child coreside. The share of children in three generation households has been increasing: In 2001, 6 % of US children lived in a three generation family household and by 2011, 8 % did (author’s calculation using CPS data).

Despite rising trends in intergenerational coresidence, research on three generation family households is relatively slim. In fact, research has shown that non-nuclear family relationships are often overlooked in family literature and that about 80% of articles on families focus on couples or parents (Fingerman & Hay, 2002). This paper seeks to add to the descriptive literature on three generation family households. As studies have found that living in a three generation household is associated with outcomes for children and families, understanding the dynamics and determinants of these complex households can inform family related research.

Bengtson argued that changes in family structure, as well as greater longevity, have led to an increased reliance on kin to perform family functions. Today, about 40% of births are to unwed mothers and about half of those births are to cohabiting couples (Ventura, 2009; Manlove, Ryan, Wildsmith & Franzetta, 2010). Using data from the Fragile Families and Child Wellbeing Study (FF), this study investigates differences by family structure in one type of kin support, three generation coresidence. If Bengtson is correct, and family structure changes have led to increased reliance on kin support, we might expect to see differences in three generation coresidence by mother’s relationship status. The oversample of nonmarital births in the FF data provides a unique opportunity to study differences in three generation coresidence between fragile families (unmarried parents and their children) and married families. Specifically, this paper documents the share of families living in three generation family households by mother’s relationship status and estimates incidence in this population over time. Second, this study is the first to examine correlates of three generation family coresidence and whether they vary by mother’s relationship status. Third, this research documents patterns of transition and duration of three generation coresidence and differences by mother’s relationship status at birth.

Literature Review and Theoretical Perspectives

Recent cross sectional estimates show that about 3.8% of households include three or more generations (US Census, 2010) and that in 2010, 7.8% of children lived in a three generation family household (Kreider & Ellis, 2011). Yet cross sectional statistics do not give a sense of incidence, or how common these household arrangements are, over time. Older studies have shown that the prevalence of three generation family households is 3-4 times higher in longitudinal data than in a cross section (Beck & Beck, 1989, 1984). Although this study cannot provide national estimates of prevalence, understanding how common three generation family household living arrangements are for fragile families as compared to married couple families is especially important as fragile families are more disadvantaged both at the birth and over time. Bengtson (2001) argued that changes in family structure have led to increased reliance between generations to perform family functions. Thus we expect that fragile families are likely to need more kin support than married families as fragile families are likely to have fewer resources.

This increased reliance on kin among fragile families is in part due to variation in economic, parental, and community resources by mother’s relationship status (McLanahan & Sandefur, 1994). As married mothers have the most resources, it is likely they will have the lowest rates of coresidence with grandparents (Beck & Beck, 1989; Tienda & Angel, 1982; Aquilino, 1990). Single mothers usually have fewer resources and need the most support from kin (Hofferth, 1984; Angel & Tienda, 1982; Jayakody, Chatters, & Taylor, 1993), whereas cohabitors fall somewhere in between (Cherlin, 2009). Married adults have also been found to have fewer intergenerational ties, and cultural norms around marriage (i.e. independence) may in part explain this difference (Sarkisian & Gerstel, 2008).

No studies have looked at correlates of three generation family households in particular, although several studies have looked at correlates of multigenerational households more broadly (Kamo, 2000; Ruggles, 2003, 2007, Choi, 2003; Cohen & Casper, 2002). This literature found that correlates fell into three broad categories: economic need, culture, and generational needs. Families with fewer economic resources (less education or lower income) may need to live in a three generation family household to combine resources and take advantage of economies of scale (Kamo, 2000; Cohen, 2002; Cohen & Casper, 2002). Cultural factors and norms may also be correlated with three generation coresidence (Hawkins & Eggebeen, 1991). Black and Hispanic families are more likely to reside with kin than White families (Angel & Tienda, 1982; Hofferth, 1984; Hogan, Hao & Parish, 1990; Pebly & Rudkin, 1999; Cohen & Casper, 2002). Families that are more familistic in orientation (e.g. Hispanics, Catholics, or those that grew up in a married family) are also more likely to coreside (Baca Zinn & Wells, 2000; Oropesa & Landale, 2004). Members of certain immigrant communities may be more likely to live in a three generation family household or if individuals have immigrated without their families, less likely. Similarly, religion may influence coresidence depending on community norms and values.

Generational needs, such as the needs of the parent generation may also influence the decision to coreside (Aquilino, 1990). Young mothers or those having their first child may be more likely to live with their own parents (Hogan, Hao & Parish, 1990; Trent & Harlan, 1994). Mothers in poor health or with a needy baby (low birth weight or disabled) may also need to coreside. Equally, the needs of the grandparent generation, such as poor physical or mental health, may also influence the decision to coreside (Choi, 2003; Cohen & Casper, 2002). Research has shown that assistance generally flows from the grandparent to the parent generation (Fingerman, Miller, Birdit, & Zarit, 2009; Grundy, 2005) thus we might expect that the needs of the parent generation to be more highly correlated with coresidence (Aquilino, 1990).

Different factors may be correlated with coresidence depending upon a mother’s relationship status as family resources, cultural meaning and norms differ for married, cohabiting and single mothers. Economic needs are likely to play a more significant role for single or cohabiting mothers than married mothers who are generally better off financially; whereas cultural factors (race, immigrant status, religion) likely play a similar role across family structures. Generational needs may play a different role depending on relationship status; for married families who are economically stable and norms of independence are stronger (Sarkisian & Gerstel, 2008), we may find that grandparent needs more strongly predict coresidence, whereas among single mothers the parent’s need may be a stronger correlate.

Research on child wellbeing has shown that the stability of households plays an important role (e.g. Wu & Martinson, 1993). Yet little is known about the stability or duration of three generation family households. Research has found that three generation households are short lived, last less than two years (Beck & Beck, 1989), and transitions are common for young mothers (Oberlander, Shebl, Madger & Black, 2009). Differences in transition patterns by mother’s relationship status at the birth may also exist; married mother’s relationships are generally more stable and three generation coresidence may be longer lasting for them than for fragile families where changes in romantic relationships are common.

Method

Data

This article used data from the Fragile Families and Child Well-being Study (FF), a study that was designed to be representative of births in cities with populations over 200,000. Births were randomly sampled in 20 US cities (in 15 states) between 1998 and 2000 with an oversample of nonmarital births (Reichman, Teitler, Garfinkel & McLanahan, 2001). Mothers and fathers were interviewed soon after the birth of the focal child and follow up interviews were conducted when the child was approximately 1, 3, 5, and 9 years old.

This study utilized the mother’s surveys as they were more complete than the father interviews and mothers were more likely to be residing with their child. The sample for the baseline survey was 4,898; 4,364 for the year 1 follow up; 4,231 for the year 3 follow up; 4,139 for the year 5 follow up, and 3,511 for the year 9 follow up. The response rate for the 1-year mother’s survey was 90%, 88% for the 3-year follow up, 87% for the 5-year follow up, and 76% for the 9-year follow up. Analyses of transitions and duration (Table 4) were restricted to mothers who were interviewed in all 5 survey waves, a sample of 2,986.

Table 4.

Incidence Over Time, Patterns of Transition, and Duration of Three Generation Family Households by Mother’s Relationship Status (N=2986)

	Total (1)	Married (2)	Cohabiting (3)	Single (4)
Panel A: Incidence over Time1 N	2986	747	1055	1184
Three generation household at least 1 wave	42.63	21.82	37.63	60.22
Panel B: Patterns of Transition 1
Never 3 generation	57.37	78.18	62.37	39.78
Consistently 3 generation	1.81	0.94	0.76	3.29
3 generation → Not 3 generation	18.03	5.22	12.13	31.42
Not 3 generation → 3 generation	4.52	4.95	4.74	4.05
Two transitions (into or out)	13.44	8.43	16.11	14.19
Three or four transitions (into or out)	4.82	2.27	3.88	7.26
Panel C: Duration 2 N	1273	163	397	713
One wave only	45.33	58.9	55.16	36.75
Multiple waves - not consecutive	19.4	14.12	16.63	22.16
Multiple waves - consecutive	35.27	26.99	28.21	41.09
Two consecutive waves	16.97	15.34	15.11	18.37
Three consecutive waves	8.33	4.29	6.8	10.1
Four consecutive waves	5.73	3.07	4.28	7.15
Five consecutive waves	4.24	4.29	2.02	5.47

¹Note: Sample includes only respondents who are in all 5 survey w aves.

²Sample includes only respondents who are in all 5 survey waves and li ved in a 3 gen eration family household in at least one survey wave.

Analyses of attrition showed that mothers who attrited were more disadvantaged than those who remained in the sample. Attriters were less educated, had lower income to needs ratios, and were more likely to be immigrants and Hispanic. To deal with the attrition, the data were multiple imputed (Allison, 2002; Rubin, 1976). Multiple imputation uses the observed data to impute values for individuals who have missing data. Five data sets were imputed (using the ice command in STATA 12) and the estimates were averaged over these data. The descriptive results were virtually identical in the observed and imputed analyses (rates of transition were about 1/2 a percentage point higher in the imputed data). To take a more conservative approach and not utilize the data that imputed the outcome of interest (three generation family household status), the unimputed results for the descriptive tables (Tables 1 and 4) are reported. In the multivariate analyses (Table 2 that describes the sample and Table 3 the regression results), the imputed data were utilized to retain respondents who had missing data on covariates. Analyses were run using listwise deletion as well and the results were substantively similar.

Table 1.

Percent of Three Generation Family Households by Age of the Child and by Mother's Relationship Status

	Birth n=4898	Age 1 n=4364	Age 3 n=4231	Age 5 n=4139	Age 9 n=3511
Total three generation
Unweighted	25.93	20.14	14.72	12.08	11.26
Weighted a	17.55	17.50	12.10	9.11	7.63c
Three generation by relationship status at baseline
Married	9.01	8.58	7.41	7.21	7.93
Cohabiting	16.77	16.14	12.54	10.42	9.65
Single b	44.81	31.13	21.42	16.60	14.71

^aNote: Weighted to be nationally repr esentative of births i n large US cities.

^bSingle includes mother’s who may be in a romantic relationship that is not coresident.

^cEstimated weights. Year 9 survey weights are not yet available.

Table 2.

Sample Characteristics by Three Generation Household and Mother’s Relationship Status at the Birth of the Child (N=4,898)

	All		Married		Cohabiting		Single
Three Generation	Yes	No	Yes	No	Yes	No	Yes	No
Economics N	1270	3628	108	1082	299	1481	864	1064
Mother’s education
Less than high school	40.33	29.66	20.55	14.67	49.93	35.14	39.35	38.74
High school degree	31.38	29.81	25.87	18.94	26.43	35.85	33.83	33.33
Some college	24.59	26.16	37.43	28.52	22.10	25.44	23.89	24.52
College and above	3.69	14.36	16.15	37.87	1.53	3.57	2.93	3.41
Grandmother’s education
Less than High School	18.01	22.74	28.26	21.81	24.84	25.27	14.31	20.06
High School	57.14	51.39	50.64	42.09	56.43	52.93	58.20	59.77
Some college	14.33	13.54	7.34	16.35	10.89	12.65	16.42	11.64
College and above	10.51	12.33	13.76	19.75	7.83	9.15	11.06	8.52
Culture
Mothers’ race/ethnicity
White non-Hispanic	13.91	25.24	34.77	45.60	17.01	19.32	11.48	10.63
Black non-Hispanic	56.81	44.21	29.36	23.75	43.63	44.78	64.90	66.83
Hispanic	25.78	26.92	26.15	23.58	35.86	33.88	20.89	20.44
Other non-Hispanic	3.50	3.63	9.72	7.07	3.50	2.02	2.73	2.10
Mother lived with both
parents at age 15	38.81	46.08	61.47	65.77	37.96	41.25	36.30	30.69
Mother is an immigrant	10.11	17.81	33.21	25.08	14.33	17.26	5.75	10.30
Grandparent is an immigrant	16.58	21.82	42.38	28.52	19.42	22.16	12.38	13.62
Religion
No religion	11.50	10.49	7.52	6.09	10.63	12.08	12.30	13.22
Protestant	55.16	49.33	39.82	42.69	49.75	47.87	58.99	59.08
Catholic	26.04	30.68	37.43	36.74	34.59	33.36	21.58	19.75
Other	7.30	9.50	15.23	14.48	5.03	6.69	7.13	7.95
Generational needs
Mother’s age at birth (years)	22.08	26.38	26.68	29.62	21.76	24.64	21.63	25.23
Mother’s first birth	55.09	32.66	33.39	35.41	46.24	33.67	60.93	28.03
Baby low birth weight	10.86	9.26	4.95	5.67	8.73	9.81	12.36	12.53
Grandparent depression	29.92	32.52	30.83	33.54	33.31	32.62	28.60	31.21
Grandparent substance abuse	33.54	34.70	35.59	28.88	35.16	38.43	32.71	35.84
Relationship status
Married	8.36	31.59	-	-	-	-	-	-
Cohabiting	24.09	40.85	-	-	-	-	-	-
Single	67.55	27.56	-	-	-	-	-	-

Note: Italicized figures represent statistically significant differences at the p<.05. All variables are from the baseline survey unless measured at a later wave but considered an unchanging characteristic.

Table 3.

Correlates of Living in a Three Generation Family Household at the Birth of the Child - Logistic Regressions (N=4898)

	Full Sample			Married			Cohabiting			Single
Economics	B	SE B	OR	B	SE B	OR	B	SE B	OR	B	SE B	OR
Mother’s education
Less than high school a, b	0.18+	(0.09)	1.19	0.17	(0.33)	1.19	0.52**	(0.16)	1.68	0.04	(0.12)	1.04
Some college	0.04	(0.10)	1.04	0.21	(0.28)	1.23	0.18	(0.19)	1.19	−0.07	(0.14)	0.93
College and above	−0.64**	(0.20)	0.53	−0.92*	(0.38)	0.40	−0.35	(0.50)	0.71	−0.12	(0.31)	0.89
Grandmother’s education
Less than High School	−0.12	(0.14)	0.89	−0.36	(0.34)	0.70	−0.03	(0.21)	0.97	−0.18	(0.18)	0.83
Some college b, c	0.01	(0.11)	1.01	−0.70+	(0.41)	0.50	−0.21	(0.22)	0.81	0.28	(0.17)	1.32
College and above b, c	0.08	(0.14)	1.08	−0.17	(0.35)	0.85	−0.18	(0.25)	0.83	0.29	(0.21)	1.33
Culture
Black non-Hispanic	0.29*	(0.12)	1.34	0.41	(0.30)	1.50	0.20	(0.20)	1.22	0.19	(0.17)	1.20
Hispanic	0.28+	(0.14)	1.33	0.35	(0.36)	1.41	0.16	(0.23)	1.18	0.23	(0.21)	1.26
Other non-Hispanic	0.75**	(0.22)	2.12	0.91+	(0.45)	2.48	0.98*	(0.39)	2.65	0.42	(0.34)	1.52
Lived with both parents at 15	0.24**	(0.08)	1.27	0.09	(0.24)	1.10	0.08	(0.14)	1.09	0.43**	(0.11)	1.54
Mother immigrant b, c	−0.60**	(0.15)	0.55	−0.39	(0.34)	0.68	−0.34	(0.25)	0.71	-1.02**	(0.24)	0.36
Grandparent immigrant a	0.19	(0.14)	1.21	0.75*	(0.32)	2.12	−0.19	(0.25)	0.83	0.23	(0.21)	1.26
Protestant	0.16	(0.13)	1.17	−0.22	(0.47)	0.80	0.18	(0.23)	1.20	0.15	(0.16)	1.17
Catholic	0.17	(0.15)	1.19	−0.35	(0.48)	0.70	0.30	(0.26)	1.35	0.17	(0.21)	1.18
Other	0.07	(0.18)	1.07	−0.18	(0.50)	0.84	0.06	(0.36)	1.06	0.08	(0.25)	1.08
Generational needs
Age : Under 18a	1.09**	(0.22)	2.97				0.72+	(0.42)	2.05	1.14**	(0.28)	3.11
18-19	0.35*	(0.15)	1.41	0.31	(0.64)	1.37	0.17	(0.25)	1.18	0.42*	(0.20)	1.53
20-24	0.12	(0.11)	1.12	0.30	(0.34)	1.35	−0.04	(0.19)	0.96	0.16	(0.15)	1.18
25-34a,b	−0.68**	(0.12)	0.51	−0.19	(0.30)	0.83	−0.70**	(0.23)	0.50	−0.77**	(0.17)	0.46
Mother’s first birth a, b, c	0.71**	(0.08)	2.03	0.11	(0.24)	1.11	0.46**	(0.14)	1.58	1.00**	(0.11)	2.72
Low birth weight baby	0.02	(0.12)	1.02	−0.15	(0.49)	0.86	−0.13	(0.23)	0.88	0.11	(0.15)	1.12
Grandparent depression	−0.03	(0.09)	0.97	−0.28	(0.26)	0.76	0.14	(0.16)	1.15	−0.07	(0.12)	0.93
Grandparent substance abuse a	−0.07	(0.09)	0.93	0.40	(0.26)	1.50	−0.24	(0.15)	0.79	−0.07	(0.12)	0.93
Relationship status
Cohabiting	0.22	(0.13)	1.25
Single	1.56**	(0.13)	4.77
Constant	−2.51**	(0.21)	0.08	−2.30**	(0.59)	0.10	−2.08**	(0.34)	0.12	−1.01**	(0.27)	0.36
Observations	4898			1190			1780			1928

Note: Standard errors in parentheses.

^aMarried-cohabiting difference in coefficie nt is sign ificant at p<.05.

^bMarried -single difference in coeffic ient is si gnificant at p<.05.

^cCohabiting-single difference in coefficient is significant at p<.05.

⁺p < .10

^*p < .05.

^**p < .01

Measures

Three Generation Family Structure

At each survey wave a measure of three generation family structure was constructed as a dummy variable set to one if a grandfather, a grandmother, or both were listed in the household roster.

Mother’s Relationship Status

Relationship status was constructed based on mother’s relationship status at the birth of the child. Mothers were coded as married, cohabiting with the baby’s father, or single. Single mothers may have been in a romantic relationship with the baby’s father (or another partner) but were not coresident. Unlike three generation family structure, this variable does not change over time; it is a measure of the relationship status at the birth. Using relationship status at birth allowed for an investigation of differences in the reliance on kin networks between fragile families and married families over time. Both cohabiting and single mothers are considered fragile families because they were unwed at the birth of the child.

Economic Need

Two measures of economic need were included in the analyses: mother’s education and grandmother’s education. Education was specified as less than high school, high school (reference), some college, and a bachelor’s degree or higher. Education was used instead of family income because income is likely to be endogenous (affected by three generation status).

Culture

Cultural factors include race, immigrant status, mother’s family background and religion. Race or ethnicity was coded as a series of dummy variables for non-Hispanic White (reference category), non-Hispanic Black, Hispanic, and other non-Hispanic race. Dichotomous variables indicating whether the mother is an immigrant and whether either grandparent is an immigrant were included. A dichotomous measure indicating whether the mother lived with both her parents at age 15 was used as a measure of family background. Religion was coded into four categories: Protestant, Catholic, other, and no religion (reference).

Generational Need

The needs of the parent generation were captured with several variables. Mother’s age, entered as a set of dummies (14-17, 18-19, 20-24, 25-35, and 35+ − reference), a dummy variable indicating whether she was having her first child, and a dummy variable indicating whether the baby was low birth weight (<2500 grams). Grandparent’s needs were captured using a measure of whether the grandparent had a drug or alcohol problem and whether the grandparent had depression when the mother was growing up.

Transition Patterns

In order to determine complete patterns of transition, this variable was restricted to respondents who were present in all survey waves (N=2,986). Patterns of transition coded respondents into 6 categories: (a) always three generation (in a three generation family household at baseline, years - 1,- 3, - 5, and - 9), (b) never three generation, (c) start in a three generation household and transition out (regardless of when the transition actually occurred), (d) start out of a three generation household and transition in, (e) two transitions in or out of a three generation family household (regardless of starting position), and (f) three or four transitions into or out of a three generation family household. Distinguishing the number of transitions captures stability. For families with a single transition (codes c and d), distinguishing the starting point (coresident or not) highlights differences between families who needed some help at the birth of the child and moved out, versus those who moved in later, likely as a result of some crisis.

Duration

In order to assess the duration of three generation family households (from baseline to year 9) respondents were coded as coresident one wave only, two nonconsecutive waves, two consecutive waves, three nonconsecutive waves, three consecutive waves, four nonconsecutive waves, four consecutive waves, and five consecutive waves.

Analytic Strategy

This study is descriptive in nature and utilizes bivariate (weighted and unweighted) statistics to document the share of families in a three generation family household and differences by mother’s relationship status at the birth of the child. In order to investigate correlates of three generation family coresidence at the birth of a child, a multivariate analysis was conducted utilizing logistic regression. Equation (1) shows the regression model:

$${\mathrm{Y}}_{\mathrm{i}}={\beta }_0+{\beta }_1{\text{Economic}}_{\mathrm{i}}+{\beta }_2{\text{Cultural}}_{\mathrm{i}}+{\beta }_3{\text{Generational Needs}}_{\mathrm{i}}+{\epsilon }_{\mathrm{i}}$$ (1)

where Y_i is the outcome of interest – living in a three generation family household at the birth of the child, and Economic, Cultural, and Generational Needs are characteristics that may be associated with living in a three generation family household. These analyses were repeated stratifying the sample by mother’s relationship status at the birth of the focal child to investigate whether the correlates of three generation coresidence varied by relationship status. Lastly, the analyses of patterns of transition and duration utilized bivariate descriptive statistics.

The FF data collection started at the birth of the child; living arrangements of families prior to the birth were not observed, therefore the independent variables were restricted to characteristics that are generally unchanging, or predate the birth to avoid issues of reverse causality. This approach has its limitations, chiefly that many of the potential covariates that might be correlated moving into a three generation family household are measured at the same time as the household structure and are therefore endogenous to the outcome of interest (like household income or employment) and are excluded from these analyses.

Results

Share and Correlates

Table 1 shows the percent of respondents who lived in a three generation family household by the age of the child. Both the unweighted and weighted percentages are reported for the overall share of three generation families. Similar to other studies, three generation coresidence was most common when children were very young (Bryson & Casper, 1999; Pebly & Rudkin, 1999; Mutchler & Baker, 2004). Nearly 26% of respondents lived with at least one grandparent at the birth of the focal child and by year 9 this had decreased to 11%. The weighted results take into account the oversample of nonmarital births to make the data nationally representative of births in large US cities. Even after weighting, nearly 18% of respondents lived in a three generation family household at the birth of the child, suggesting many urban families rely on grandparent support at the time of the birth of a new child.

As anticipated, the frequency of living in a three-generation household varied greatly by the relationship status of the mother. Among mothers who were married at the birth, 9% lived in a three generation family household at the birth of the child; this share decreased over time to about 7%. Interestingly, the share of married mothers living in a three generation family household increased between age 5 and 9 by about 1 percentage point. The year 9 data were collected from 2007-2010 and coincided with the Great Recession. The increase among married mothers may reflect increased doubling up due to the economic crisis. Among cohabiting mothers, the share that lived in a three generation family household at the birth of the child was nearly double (17%) that of married mothers, and by age 9, it was still higher than for married mothers at nearly 10%. Lastly, as expected, a much higher share – 45% – of single mothers lived in a three generation family household at the birth of the focal child but this dropped to about 15% by age 9. These findings suggest that among mothers who were married at the birth, their reliance on kin was somewhat stable over time, whereas for fragile families, kin support was particularly important when they had a very young child.

Table 2 reports the sample characteristics at the birth of the child. Similar to prior literature, economic, cultural, and generational needs differed for individuals in three generation family households and those who were not. Mothers and grandmothers in three generation family households had lower levels of education than those not in three generation family households. Cultural factors also differed; mothers in three generation family households were less likely to be White, more likely to be Black, and less likely to have lived with both parents at age 15. Lastly, in terms of generational needs, mothers in three generation family households were significantly younger and more likely to have had a first birth than other mothers. Grandparents in three generation family households were less likely to have been depressed than those not in a three generation family household. Coresident mothers were also more likely to be single and less likely to be married.

Once the sample was divided by mother’s relationship status, some differences emerged. In fact, the only characteristic that was statistically different between mothers in three generation family households versus those who are not across all relationship statuses was mothers’ age; mothers who resided in a three generation family household were younger than those who did not. To more fully investigate what characteristics were associated with three generation family households and how these differed by mother’s relationship status, Table 3 reports the results of logistic regressions predicting three generation coresidence at the birth of the child by mother’s relationship status.

Table 3 reports the B, the standard error on B, and the odds ratio for logistic regressions predicting three generation family coresidence. Discussion is focused on the odds ratios that indicate whether a particular characteristic is associated with a greater or lesser likelihood of living in a particular type of household as compared to the reference category. To test whether differences by relationship status were significantly different from each other, Chow tests were conducted on the fully interacted models comparing each group (married versus single, married versus cohabiting, cohabiting versus single) and significant differences are noted in Table 3.

In the full sample, economic needs of the mother predicted coresidence (mother’s lower education is marginally associated with increased odds and higher education is significantly associated with decreased odds), but grandmother’s education did not. Several cultural correlates were associated with three generation family households. As has been found in studies of multigenerational households generally, mothers who were Black, Hispanic, or other race or ethnicity were all significantly more likely to reside in a three generation family household than Whites. Mothers who lived with both parents at age 15 were more likely to coreside whereas immigrants were less likely than their peers. In terms of generational needs, the needs of the parent generation mattered, whereas those of the grandparent generation did not: younger mothers (especially teen mothers) were more likely to live in a three generation family household whereas older (25-34) ones were less likely. Mothers who had a first birth were two times as likely to coreside. In comparison with married mothers, single mothers were 4.7 times as likely to have lived in a three generation family household.

Stratifying the sample by mother’s relationship status revealed a few different patterns. As predicted, among married mothers, economic need did not play a very significant role in determining coresidence. In comparison, among cohabiting mothers, economic factors played a significant role in predicting coresidence; cohabiting mothers with less than a high school education were 68% more likely to live in a three generation family household than those with a high school education. Counter to expectation, economic factors did not play a role among single mothers. It may be that the economic factors investigated here only capture one part of economic need and other factors not available in these data are predictive of coresidence.

Differences in cultural correlates by mother’s relationship status were not expected, yet some differences emerged. Interestingly, race or ethnicity was not associated with coresidence for any of the groups except for mothers who were “other” race once the analyses were stratified but differences between relationship groups were not significant. Among married mothers, those who had parents (grandparents) who were immigrants were 2 times as likely to live in a three generation family household, yet this was not associated with coresidence for cohabiting mothers. For single mothers being an immigrant was negatively associated with three generation coresidence.

Generational needs of the parent generation were predictive of coresidence, but not those of the grandparent generation (this may be in part due to a lack of sufficient measures of grandparent needs). Older mothers who were cohabiting or single were less likely to live in a three generation family household whereas this was not the case for married mothers. Across relationship categories mothers who were having a first birth were all more likely to live in a three generation family household but differences in the magnitude varied by mother’s relationship status: Single mothers who were having a first birth were two and a half times as likely to coreside, cohabiting mothers 58% more likely, and married mothers only 11% more likely. Thus, it appears that the needs of the parent generation predict coresidence more strongly than the needs of the grandparent generation.

Patterns of Transition and Duration

To fully understand the complexity of three generation family households and the use of kin support among fragile as compared to married families it is important to look at incidence over time, transition patterns, and duration. Figure 1 illustrates the patterns of transitions into, and out of, three generation family households. Mothers were plotted as three generation (left) or not (right) at baseline and their transitions were plotted over time (moving down). In order to provide complete information on transitions, the sample was restricted to only mothers who were interviewed in all five survey waves. The widths of the lines connecting the transitions were drawn to approximately represent their relative sample sizes.

Figure. 1.

Flow Chart of Mothers in a Three Generation Family Household over 5 Survey Waves (N=2986)

Note: Sample restricted to mothers in all survey waves.

Figure 1 demonstrates the fluidity of three generation household arrangements and the volume of transitions into and out of three generation family households among this population. Strikingly, nearly 43% of this sample (and 40% of the total sample) lived in a three generation family household in at least one of the 5 survey waves. Although 26% of respondents coresided with a grandparent at the birth of the child and this percentage decreased over time, in fact many more families relied on kin for support at some point in time.

Figure 1 plots the transitions for all mothers regardless of relationship status; Table 4 displays patterns of transition by mother’s relationship status. Panel A of Table 4 shows the percentage of mothers who ever reported living in a three generation family household by mother’s relationship status. Again, the sample was restricted to those who were interviewed in all waves of the survey. Sixty-percent of mothers who were single at the birth coresided with a grandparent in at least one wave, 38% of cohabiting mothers, and only 22% of married mothers. Kin support was important across family types and was most important for fragile families.

Panel B of Table 4 documents patterns of transition by relationship status. Very few mothers lived in a three generation family household consistently over the 5 survey waves (1.8%) and single mothers were by far the most likely to do so (3%). Starting out in a three generation family household and then leaving that household was the most common pattern observed in these data. This pattern corresponds to the idea that early on in a child’s life, families may require additional assistance from kin, but they later move out of the three generation family household and remain outside of the household. Single mothers were by far the most likely to follow this pattern (31%) and 12% of cohabiters did the same, but only 5% of married mothers. Starting outside of a three generation household and then moving into one implies a different mechanism. Mothers who moved in may have experienced a shock of some sort that led them to move in with kin. Thus, we did not expect to see differences in this pattern across mother’s relationship status and this was the case with 4-5% following this pattern. Eighteen percent of the sample made two or more transitions into (or out of) a three generation family household. Fragile families were expected to make the most transitions as their relationship statuses are more likely to change over time and as their financial situation may be more precarious than married mothers, and this was the case. Eleven percent of married mothers made two or more transitions into or out of a three generation family household, whereas 20% of cohabiting or single mothers did the same. Single mothers had the most transitions with 7% of the sample making 3 or 4 transitions.

Panel C of Table 4 reports the number of consecutive (and not consecutive) waves of coresidence to estimate the duration of three generation households. Among mothers who ever reported living in a three generation family household, 45% only reported doing so in one survey wave. Again there were differences by mother’s relationship status but not necessarily as expected. Single mothers were the least likely to coreside only for one wave, whereas 59% of married mothers only coresided one wave. In terms of the number of consecutive waves of coresidence, none of the differences by relationship status were statistically significant although single mothers overall were more likely to reside in a three generation family household consecutively.

Conclusion

Bengtson argued that multigenerational bonds are becoming more important for the wellbeing of families and as a source of support. The findings of this study certainly support the assertion that multigenerational support is widespread. Close to half (43%) of families in the study lived in a three generation family household in at least one survey wave and this finding likely underestimates the true incidence among this population as there were long intervals between survey waves. Fragile families (unmarried parents and their children) were most likely to rely on coresidence with kin both at the birth and over time. Sixty-percent of single mothers in the study lived in a three generation family household at least once over the first 9 years of the focal child’s life. The need for intergenerational support also appears to be higher when children are younger. At the birth of the child 26% of the sample resided in a three generation family household (18% of the weighted sample) and only 11% did likewise at age 9. Together, these statistics suggest that urban families and fragile families in particular, rely a great deal on intergenerational support through coresidence.

Mother’s relationship status was associated with three generation family coresidence. Moreover, the factors correlated with coresidence varied by mother’s relationship status. Economic factors were not correlated with coresidence among single mothers but were associated with coresidence for cohabiting mothers. Cultural factors associated with coresidence also varied by mother’s relationship status. As anticipated, race or ethnicity was correlated with coresidence similarly across relationship groups, but for single mothers being an immigrant herself was associated with decreased odds.

Regardless of mother’s relationship status, the needs of the parent generation appeared to be more strongly associated with coresidence than the needs of the grandparent generation (although that may in part be due to incomplete information on potential needs – such as the health of the grandparent generation). Mothers who were younger were more likely to coreside with a grandparent, and across all relationship statuses, mothers who had a first birth were more likely to coreside although single mothers with a first birth were the most likely.

Lastly, this study investigated the patterns of transition and duration of three generation family households. Families transitioned into and out of three generation family households very often (18% transition two or more times) and transitions were more common among fragile families. Three generation family households were generally short-lived. Among the families who lived in a three generation family household, nearly half did so only at one survey wave. Although the tenure of three generation households was short, given the frequency of their occurrence, and their likelihood of reoccurrence, these households likely play an important role in the lives of children, mothers, and grandparents. Other research has documented the importance of returning home for adult children, in particular middle class children (Newman, 2012), but these findings suggest that families also provide an important safety net for adult children with kids, and among the economically disadvantaged as well. As Bengtson suggested, multigenerational support is important.

This study has several limitations. First, the sample while focused on urban, primarily low income mothers, a population that is very likely to live in a three generation family household, is not nationally representative. Future studies should utilize nationally representative data to see if these differences between fragile families and married families hold. Second, although the analyses of the correlates of three generation family coresidence are suggestive, FF data have only limited information on the families prior to coresidence. Future research that looks at predictors of three generation families would ideally have more information on families prior to coresidence and additional variables that capture economic and generational needs (in particular grandparent needs such as physical health). Lastly, the data were collected periodically and cannot capture three generation family coresidence between waves; therefore it is likely that the percentages presented here are underestimates.

Despite these limitations, the findings from this study are suggestive for future research and policy making. Many families today, especially when they have a very young child, live in a three generation family household and rely on kin support throughout early childhood. Three generation family households appear to be very transitory and research on nuclear family transitions suggests that instability is not good for children, but this may not be the case at the multigenerational family level. Future research should consider transitions not only at the nuclear family level but also at the three generation family level. Moving in with grandparents may provide stability (economically or emotionally) in times of crisis or perhaps, as is the case with nuclear family transitions, these disruptions are detrimental to child development. Policies are often targeted at the household level and recognizing the complexity of households is important.

This study is the first to investigate correlates of living in a three generation family household and patterns of transitions and duration of these households. It is also the first to examine differences in three generation family households by mother’s relationship status comparing fragile families to married families. This study sheds light on the complexity of this increasingly common household arrangement, one that is likely to become more important in years to come, as the population ages and as out of wedlock birth rates remain steady or rise.