Fathers’ Investments of Money and Time Across Residential Contexts

Abstract

Fathers’ roles in family life have changed dramatically over the past 50 years. In addition to ongoing breadwinning responsibilities, many fathers are now involved in direct caregiving and engagement with children. Yet there is considerable variation in what fathers do, especially depending on whether they live with or away from their child. In this article, the authors use data from the Fragile Families and Child Wellbeing Study (N = 3,869) to describe how fathers’ economic capacities (money) and direct involvement with children (time) are associated over child ages 1 to 9 for resident versus nonresident fathers, net of confounding factors. They found suggestive evidence that money and time investments operate differently across residential contexts: Resident fathers experience a trade-off between market work and time involved with children. In contrast, nonresident fathers’ higher economic capacities are associated with more time involvement, underscoring the greater challenge for such fathers to remain actively involved.

Over the past half-century, major demographic changes have fundamentally altered fathers’ roles in family life. In particular, union dissolution has increased, owing to high divorce rates and the growing proportion of births that occur outside of marriage (which are less often followed by long-term union stability than are marital births). Because children are more likely to live with their mother when their parents’ union ends (Amato, 2000), fathers today are more likely to be living away from their biological children than are mothers—and than their mid-20th-century counterparts. Overall, a striking aggregate decline in the proportion of U.S. men living with their own biological children has been observed between the mid-1960s and the mid-1990s (Eggebeen, 2002) and likely beyond. From the perspective of children, according to recent census data, 27% of all children under age 18 were living away from their biological father in 2014 (U.S. Census Bureau, 2014).

In addition to changing demographic patterns, fathers’ involvement in family life has also changed substantially over the past half-century, reflecting shifts in family gender roles and the division of household labor and market work. Although providing economic support remains important, fathering today also often includes nurturing and providing direct care; engaging in cognitive stimulation and play activities; taking responsibility for coordinating children’s care and activities; and connecting the child to his or her extended family, community members, and other resources (Cabrera, Tamis-LeMonda, Bradley, Hofferth, & Lamb, 2000; Marsiglio, Amato, Day, & Lamb, 2000; Palkovitz, 2002).

At the intersection of changes in family demography and the nature of fathering is the fact that there is, today, great heterogeneity in how men enact the father role. Although a significant body of research has explored the “new fatherhood” and its consequences for children (Cabrera & Tamis-LeMonda, 2012; Lamb, 2010), less well understood is how the various dimensions of contemporary fathers’ involvement may be interrelated, particularly across residential contexts. In this article we provide descriptive evidence about how fathers’ economic capacities and contributions are linked to fathers’ time involvement (across several measures), comparing resident versus nonresident fathers over Years 1–9 after a focal child’s birth. This is an important question because this interrelationship may affect the total amount of investment that children “get,” which is linked with children’s well-being and development (Bornstein, 2006; Lamb, 2010). Also, this work highlights the challenges for nonresident fathers to remain actively involved with children, as they cannot benefit from the joint allocation of parenting tasks that occurs among coresident couples and as they must navigate greater complexity when new partners (of fathers or mothers) get involved (Tach, Mincy, & Edin, 2010).

Conceptual Framing and Prior Empirical Research

Family scholars often point to two key domains of parental investments in children—economic resources and parenting behaviors—or, colloquially, money and time (Thomson & McLanahan, 2012). Economic resources enable parents to provide the food, clothing, and shelter requisite for daily living as well as the material goods and experiences that promote positive child development (Duncan & Brooks-Gunn, 1997; Magnuson & Votruba-Drzal, 2009). Parenting time encompasses the wide range of interactions, behaviors, and activities that allow parents to provide the appropriate warmth, support, control, and monitoring that are intrinsic to high-quality (or so-called “authoritative”) parenting, which has been shown to be positively related to children’s well-being (Amato & Fowler, 2002; Baumrind, 1986).

Although early research on parenting emphasized the importance of mothers for children, the past few decades have produced a large literature more specifically focused on fatherhood, exploring variation in its content, context, and implications (Furstenberg, 1988; Lamb, 2010; Marsiglio et al., 2000). As noted above, the father role has broadened from the heyday of the breadwinner–homemaker family model in the mid-20th century (Cherlin, 2009), yet providing economic support remains a key element of fathering and fathers’ identity today (Christiansen & Palkovitz, 2001). Economic contributions and capacities are reflected in the work hours that fathers expend in the labor market, their earnings (which result both from work hours and reflect characteristics such as age and education), and—for nonresident fathers—the actual amount that they pay toward children’s expenses in the form of child support. With respect to fathers’(noneconomic) involvement with children, one of the first and most enduring typologies of father involvement identified three key components: (a) accessibility, (b) engagement, and (c) responsibility (Lamb, 2010; Lamb, Pleck, Charnov, & Levine, 1985). Accessibility refers to time fathers are available to children (even if not directly interacting), engagement refers to fathers’ time with children doing activities together, and responsibility refers to fathers’ helping to arrange resources and activities for children (Lamb, 1986). We explore each of these here.

Links Between Fathers’ Money and Time Investments, by Residential Status

For cohabiting and, especially, married men, paternal investments of money and time occur conjointly within what some scholars have called the “package deal” (Furstenberg & Cherlin, 1991; Townsend, 2002). Becker’s (1965) classic theory of household production posits that households “combine time and market goods to produce more basic commodities that directly enter their utility functions” (p. 495). As Weiss and Willis (1985) argued, children are a collective good, and parents living together (especially with the insurance value of marriage) can jointly allocate their investments of time in market work and time with children to achieve an efficient outcome (Steinberg, 2001). Because time is finite, time spent in the labor market is by definition time not spent with children, so we would expect a negative relationship between employment and parental involvement generally, but especially for fathers, who have historically spent more time in breadwinning (Coltrane, 1996; Knoester & Eggebeen, 2006).

By contrast, nonresident fathers face both barriers and disincentives to their involvement with children, and providing economic support may in fact facilitate fathers being involved in other ways. When fathers live apart from their children, they have less incentive to invest in them because they cannot be certain how any contributions will be spent (Weiss & Willis, 1985). When combined with the lost economies of scale (two households vs. one), child-rearing investments by nonresident fathers are likely to be inefficiently low without policy intervention (Willis, 2000). Also, mothers may serve as “gatekeepers” to children after separation (Allen & Hawkins, 1999), and financial contributions help mothers to perceive fathers as more competent (Fagan & Barnett, 2003). In particular, fathers’ greater work effort, higher earnings—and, even more so, their actual financial contributions—may give them (and the mother) a greater sense of being a “good” father and hence encourage their involvement in other ways (Johnson, Levine, & Doolittle, 1999; Lerman & Sorenson, 2000). Recent studies have found a positive link between paternal employment of unmarried, nonresident fathers and father–child contact (Cabrera, Fagan, & Farrie, 2008; Ryan, Kalil, & Ziol-Guest, 2008) as well as between fathers’ child support (especially informal support) and fathers’ time with children (Amato, Meyers, & Emery, 2009; Garasky, Stewart, Gundersen, & Lohman, 2010; Huang, 2009; Nepomnyaschy, 2007). Therefore, although, as for resident fathers, time at work cannot be time spent with children, we would expect a positive relationship between nonresident fathers’ economic capacities and time involvement.

Confounding Factors

Selection may play a role in how fathers’ economic characteristics are associated with fathers’ time involvement with children. Disadvantaged fathers (who are often young, racial/ethnic minorities, had a nonmarital birth, have been incarcerated, had children with more than one partner, and experienced a nontraditional family structure in youth) generally have weaker connections to the formal economy and lower levels of father involvement than their more advantaged counterparts (Berger & Langton, 2011; Cheadle, Amato, & King, 2010). Additional factors that may influence father involvement include health (Coley & Hernandez, 2006), religiosity (King, 2003), and fathering attitudes and context (Carlson & Berger, 2013; Waller, 2012). The gender of the child can influence fathers’ work effort and time involvement (Lundberg, McLanahan, & Rose, 2007), and a grandmother living in the home provides another set of hands to help with child rearing but may also alter fathers’ duties and responsibilities (Dunifon, Ziol-Guest, & Kopko, 2014).

The Present Study

In this article we provide new information about fathers’ involvement with children over Years 1–9 after a focal child’s birth, using data from the Fragile Families and Child Wellbeing Study (http://www.fragilefamilies.princeton.edu).

We examine how fathers’ economic capacities and characteristics (money) are related to their direct involvement (time) with children, exploring differences between resident and nonresident fathers. This research sheds light on the nature of contemporary fatherhood and what children “get” from parents and highlights the challenges for nonresident fathers to stay involved.

Method

Data

We use data from the Fragile Families and Child Wellbeing Study, a longitudinal study of urban births (with an oversample of nonmarital births) that occurred between 1998 and 2000. The study includes 4,897 births (3,710 unmarried and 1,187 married), and the weighted sample is representative of births in U.S. cities with populations over 200,000. Baseline interviews with mothers and fathers took place in 75 hospitals in 20 cities (in 15 states) just after the baby’s birth, and follow-up interviews were conducted about 1, 3, 5, and 9 years after the birth. Overall, 4,331 fathers were interviewed at least once from birth through Year 9. At the 1-, 3-, 5-, and 9-year follow-up surveys, the proportions of eligible unmarried (at birth) fathers interviewed were 71%, 69%, 67%, and 56% and of married (at birth) fathers were 82%, 82%, 78%, and 69%, respectively.

Analytic Sample

We limited our sample to fathers who were interviewed at least once over the 1-, 3-, 5-, and 9-year survey waves (N = 4,092). We then dropped fathers who were interviewed in jail (because incarceration hinders fathers’ ability to contribute both time and money; n = 102, 2.5%) and cases in which the focal child did not reside with the biological mother at least half-time (n = 29, 0.7%). Also, 92 observations (2.2%) were dropped because of missing data on the main variables of interest (father’s economic capacities or time involvement). The full analytic sample for this article consists of 3,869 fathers. We pooled interviewed cases across the four survey waves, yielding 11,055 person-year observations, with a mean number of 2.9 survey waves for each father. Fathers’ residence status relative to the focal child is reported at each survey wave and can change across waves (if the father moves in or out); resident fathers contributed 7,167 person-year observations, and nonresident fathers contributed 3,888 person-year observations.

Observations with missing data on covariates were multiply imputed with chained equations using the mi commands in Stata 14. All variables were included in the imputation equations, and 10 imputed data sets were created. The fraction missing was 11% or less for every variable except earnings, with 30% of person-years missing primarily because of missing data at the 1-year survey (we imputed this information as noted later). Only observations with complete information on the outcome variables were used for analyses (White, Royston, & Wood, 2011). The sample varies slightly across the estimated models, as the number of cases with complete information on each father involvement measure fluctuates. It is important to keep the analytic sample in mind when interpreting the findings; fathers lost to attrition by the 9-year survey were more disadvantaged than those who remained.

Fathers’ Economic Characteristics (Money)

We used several measures of fathers’ economic characteristics in order to capture different aspects of fathers’ work effort and capacity as providers. Across the 1-, 3-, 5-, and 9-year surveys fathers reported on their employment and earnings during the previous year. For our employment measure, we combined hours worked per week and weeks worked per year to create a measure of average weekly hours worked. For our earnings measure, we used information based on fathers’ reports about earnings from all jobs in the past year to get a measure of annual earnings (converted to 2008 dollars—the approximate year of the last survey wave). Fathers who said they “didn’t know” their exact earnings were given a range of values, which we then converted to median values within categories. In addition, a direct question about total earnings from all jobs was not asked of all respondents at the 1-year survey, so we imputed the missing observations with a series of predictors: currently employed, weeks worked per year, relationship status, race, education, immigrant status, and ever incarcerated.

For nonresident fathers, in addition to work hours and earnings, we also include the total annual amount of child support paid, reported by mothers. At each survey wave, mothers indicated the amount of formal child support received in the past year (or since a formal order began or changed) as well as the amount of informal support received. We summed these amounts and converted the total into an annualized metric. Nonresident fathers without a formal order and no informal agreement with the mother were coded as 0. Fathers’ child support payments are inflation-adjusted and reported in 2008 dollars.

Fathers’ Direct Involvement (Time)

We measured fathers’ direct involvement with children with multiple variables that correspond to the key domains of paternal involvement: accessibility (time), engagement, and responsibility. We relied on mothers’ reports of fathers’ involvement in order to avoid inflated associations due to shared method variance, whereby the same individual reports on both the independent and dependent variables (Marsiglio et al., 2000). Fathers’ accessibility to (or time with) the child reflects how often the father spent one or more hours a day with the child in the past month, ranging from 1 (never) to 5 (every day). We measured paternal engagement in father–child activities at Years 1, 3, and 5 as the mean number of days (0–7) that the father did each of four activities with the child in the past week: singing, reading stories, telling stories, and playing with toys (αs = .91, .90, .91, respectively). At Year 9, engagement included different activities reflecting children’s older age: playing sports or outdoor activities, reading with or talking about books, talking with the child about his or her day, and helping with homework. Also, the response scale was changed to reflect activities in the past month, ranging from 1 (never) to 5 (every day); we assigned the 1–5 scores to 0–7 days to yield a range of variation similar to those at prior waves (α = .85). To examine fathers’ shared responsibility for child-related tasks, we relied on mothers’ reports about how often the father (a) “Looks after [child] when you need to do things,” (b) “Runs errands like picking things up from the store,” and (c) “Takes [child] places (he/she) needs to go, such as to day care or the doctor.” Response choices ranged from never (1) to often (4), with higher average scores indicating greater shared responsibility (α = .89–.95 across waves). An additional measure of fathers’ time with children was available for nonresident fathers: Mothers reported the number of days fathers saw the child in the past month (range: 0–30). We also created a composite involvement measure that used all items from the time, engagement, and responsibility measures common to nonresident and resident fathers; each item was standardized with a mean of 0 and a standard deviation of 1 (α = .76–.77 for resident fathers and .91–.94 for nonresident fathers across waves).

Covariates

We included a number of time-constant and time-varying covariates that may be associated with both fathers’ economic capacities and time involvement. All time-constant measures were reported by fathers at the baseline survey unless otherwise noted. We include whether the parents were married when the focal child was born (reported by mothers). Fathers’ age at birth was self-reported in years. Fathers’ race/ethnicity was specified as non-Hispanic White (reference), non-Hispanic Black, Hispanic, and non-Hispanic other race. Fathers’ education was measured with categories of less than high school (reference), high school degree, some college, and bachelor’s degree or higher. Immigrant status indicated that the father was born outside the United States. Fathers’ family background reflected whether he lived with both biological parents at age 15. We also included fathers’ attitudes toward fathering, based on whether they responded strongly disagree (1) to strongly agree (4) with the following statements: (a) “Being a father and raising children is one of the most fulfilling experiences a man can have,” (b) “I want people to know that I have a new child,” and (c) “Not being a part of my child’s life would be one the worst things that could happen to me.” Responses were averaged, and higher scores indicated more positive attitudes (α = .73). We also indicated whether the child was a boy.

Time-varying covariates were measured at each wave and were reported by fathers unless noted otherwise. Fathers’ self-reported health status ranged from poor (1) to excellent (5). Religious service attendance ranged from never (1) to more than once a week (5). Fathers with a major depressive episode were classified according to the Composite International Diagnostic Interview—Short Form (Kessler, Andrews, Mroczek, Ustun, & Wittchen, 1998). Mothers’ and fathers’ combined reports indicated whether the father had ever been incarcerated by each follow-up wave. The father reported the total number of biological children he had had with the focal child’s mother and with other women (multipartnered fertility).

We also included time-varying covariates that captured mothers’ and fathers’ relationship status. For coresident fathers, we included a variable indicating that they were legally married to the child’s mother (vs. cohabiting). For nonresident fathers, we included a dummy variable indicating they were in a dating relationship with the mother (vs. not romantically involved), and we included separate variables indicating that the father—or the mother—was involved in a romantic relationship with a new partner. Last, we included whether the child’s grandmother was living with the mother and child and the number of hours that the mother worked per week, both reported by mothers.

Sample Description

Table 1 shows the time-invariant characteristics of our sample by residence status at the 1-year survey, weighted by city sampling weights. Overall, resident fathers were more advantaged than nonresident fathers. Resident fathers were more likely to have been married to the child’s mother, to have higher levels of education, to have lived with both biological parents at age 15, and to have more positive fathering attitudes. Nonresident fathers were much more likely to be of minority race/ethnicity, to have been incarcerated, and to have had a child with another partner.

Table 1.

Sample Descriptives by Fathers’ Coresidence Status at 1 Year

	Coresidence status
	Resident		Nonresident
Background variables	M or %	SD	M or %	SD
Father characteristics
Married at child’s birth (mother report)	66.6		7.0a
Age at child’s birth (M)	30.60	6.31	25.81a	9.48
Race
White, non-Hispanic	32.5		9.4a
Black, non-Hispanic	27.8		64.4a
Hispanic	34.1		20.1a
Other, non-Hispanic	5.6		6.1
Education
Less than high school	24.2		35.0
High school diploma or GED	24.7		42.0a
Some college	24.0		20.6
BA or higher	27.1		2.5a
Foreign born	31.5		10.6a
Lived with both parents at 15	61.9		36.6a
Physical health (M, range: 1–5)b	4.01	0.89	4.05	1.44
Religious service attendance (M, range: 1-5)b	3.44	1.32	2.98a	1.98
Major depressive episode (CIDI–SF)b	7.6		19.2a
Ever incarcerated (mother report)b	12.9		34.4a
Positive fathering attitudes (M, range: 1–4)	3.78	0.36	3.54a	0.82
Other characteristics
Baby is a boy (mother report)	55.0		50.3
Grandmother lived with focal child (mother report)b	11.1		33.2a
Unweighted number of cases (n)	2,307		801

Note. All data are based on fathers’ reports, unless otherwise indicated. All figures are weighted by 1-year city sampling weights. The total number of unweighted cases here (N = 3,108) is lower than the total number of fathers in the analysis (N = 3,869) because not all fathers were interviewed at Year 1 and are thus are missing 1-year residence status. CIDI–SF = Composite International Diagnostic Interview—Short Form.

^aSignificantly different than resident fathers at p < .05.

^bTime-varying covariates. Sample statistics measured at the Year 1 survey.

As shown in Table 2, the composition of resident and nonresident fathers changed over time, as an increasing number of coresident couples broke up. As a result, changes in mean levels of father involvement across waves reflect both changes in the parenting behavior of fathers who remained in a given status as well as changes in the sample composition as fathers moved from resident to nonresident status (and a small number moved from nonresident to resident status). At 1 year, 71% of resident fathers were married to the focal child’s mother. By 9 years, this fraction had increased to 88%, reflecting marriage among some cohabiting couples and that a greater share of cohabiting than married parents broke up over time. For nonresident fathers, 18% were dating the focal child’s mother at 1 year, compared to 10% by 9 years. At 1 year, 32% of nonresident fathers were involved with a new partner, which then increased over time.

Table 2.

Means of Fathers’ Economic Characteristics, Fathers’ Involvement, and Time-Varying Controls, by Fathers’ Coresidence Status at Survey

	Year 1		Year 3		Year 5		Year 9
Variable	M or %	SD	M or %	SD	M or %	SD	M or %	SD
Resident fathers (%)	85.3		80.7		71.9		70.7
Fathers’ economic characteristics
Annual earnings ($, in 2008)	$38.6 K	$20.0 K	$50.0 K	$43.2K	$56.0 K	$47.7 K	$55.8 K	$41.8 K
Average weekly hours worked	41.58	16.29	41.91	17.39	42.53	14.77	38.25	14.75
Did not work	4.48		7.18		5.09		7.22
Fathers’ involvement (mother’s report)
Spent 1+ hours	4.80	0.51	4.77	0.53	4.80	0.43	4.77	0.47
Engagement	3.88	1.72	3.77	1.83	3.30	1.53	3.39	1.23
Responsibility	3.46	0.52	3.50	0.52	3.56	0.48	3.51	0.46
Time-varying controls (mother’s report)
Relationship with child’s mother (%)
Married	70.8		77.5		82.5		87.5
Cohabiting	29.1		20.4		17.6		12.5
Hours per week worked by mother (mother report)	23.74	18.81	24.35	19.22	23.16	17.77	24.83	14.35
Number of children with child’s biological mother	1.73	0.92	2.16	1.03	2.37	1.05	2.76	1.07
Number of children with other partner	0.35	0.74	0.31	0.75	0.28	0.70	0.22	0.57
Number of waves as resident father	1.88	0.30	2.82	0.41	3.69	0.59	4.53	0.71
Unweighted number of resident cases (n)	2,309		1,999		1,650		1,221
Nonresident fathers (%)	14.7		19.3		28.1		29.3
Fathers’ economic characteristics
Annual earnings ($, in 2008)	$23.0Ka	$22.6 K	$19.3Ka	$25.9 K	$24.0 Ka	$31.0 K	$26.1Ka	$30.3 K
Average weekly hours worked	35.43a	29.34	32.32a	32.26	33.35a	27.06	32.93	26.23
Did not work	14.1a		22.7a		15.7a		19.1a
Annualized child support payment ($, in 2008)	$1.5K	$3.5 K	$1.4K	$3.5 K	$1.4K	$3.1	$2.3 K	$4.4 K
No child support payments	36.9		44.4		48.8		38.2
Fathers’ involvement (mother’s report)
Spent 1+hours	3.16a	2.12	2.92a	2.05	2.78a	1.84	2.50a	1.57
Engagement	1.68a	2.53	1.44a	2.21	1.00a	1.74	1.10a	1.58
Responsibility	2.23a	1.41	2.10a	1.43	1.90a	1.15	1.79a	1.03
Days with child	12.52	16.21	10.40	14.81	8.52	11.99	7.14	9.79
Time-varying controls (mother’s report)
Relationship with child’s mother (%)
Dating	17.6		7.9		5.4		9.7
Nonromantic	82.4		92.1		94.6		90.3
Father had new partner	32.4		44.2		49.7		38.1
Mother had new partner	31.1		40.1		42.9		53.9
Hours per week worked by mother (mother report)	30.32a	26.69	31.75a	26.67	30.46	25.12	32.96a	20.71
Number of children with child’s biological mother	1.42a	1.08	1.53a	1.17	1.89a	1.38	1.83a	1.30
Number of children with other partner	0.86a	1.67	1.13a	2.00	1.043a	1.53	1.13	1.59
Number of waves as nonresident father	1.70	0.62	2.21	1.01	2.57	1.37	3.00	1.55
Unweighted number of nonresident cases (n)	805		967		1,103		1,035
Overall number of unweighted cases (n)	3,114		2,966		2,753		2,256

Note. Data are based on fathers’ reports, unless otherwise indicated. All figures are weighted by year-specific city sampling weights.

^aSignificantly different than resident fathers at p < .05.

Analytic Strategy

We first present descriptive statistics on the economic and involvement variables, by residence status at each wave. Then we use two multivariate analytic techniques (with pooled data across the 1-, 3-, 5-, and 9-year surveys) to examine how fathers’ economic resources were associated with fathers’ direct involvement (time, engagement, and shared responsibility). Because we have repeated measures about the same men over time, we used models designed for panel data. Random effects models use variation both between and within fathers, including time-constant and time-varying variables, and fixed effects models use variation only within fathers over time to examine how changes in economic characteristics are associated with changes in direct involvement, controlling for time-varying variables (time-constant variables are automatically dropped). The latter reduces bias by controlling for unobserved individual, time-invariant characteristics that may be associated with fathers’ economic resources and involvement (Greene, 2003; Snijders, 2005). For example, some fathers may have personal values (e.g., conscientiousness) that predict both their hard work in the labor market and their active involvement with children. The Hausman test (at p < .05) indicated that random effects models were more efficient for resident fathers but that fixed effects models should be used for nonresident fathers (in other words, the error term is correlated with the predictor variables). Because we are conceptually interested in between-father as well as within-father variation, we present both random and fixed effects results for both resident and nonresident fathers. We estimated separate models for resident and nonresident fathers, pooling across waves based on residence status at each wave. (Note that 43% of ever-nonresident fathers and 27% of ever-resident fathers were observed only once in their respective residence statuses so do not contribute to the fixed effects estimates.) Given high correlations among our measures of economic resources and involvement (all pairwise correlations are over .6), we estimated separate models for each economic characteristic predicting each involvement measure.

Results

We first describe patterns of fathers’ economic capacities and direct involvement with children by residence status at 1, 3, 5, and 9 years following the birth of the focal child (see Table 2). As expected, we found that resident fathers had much higher economic capacities and were much more involved than nonresident fathers; in other words, children who lived with their fathers received greater paternal investments of both money and time. Levels of resident paternal involvement appeared to be quite stable over time. By contrast, nonresident fathers’ average earnings and child support payments increased across waves, partly because formerly resident fathers entered this group after union dissolution, whereas paternal time involvement tended to decrease across waves. Because, as noted above, the sample of nonresident fathers becomes more positively selected over time, the decline we observe is likely underestimated.

We now turn to our multivariate results about how fathers’ economic characteristics and time involvement are linked. Table 3 shows random effects and fixed effects results for resident and nonresident fathers. We remind readers that for each father-involvement measure (listed down the left column as Panels A–E) a separate regression was estimated for each economic variable and is presented in an individual cell.

Table 3.

Fathers’ Economic Characteristics as Predictors of Fathers’ Involvement, by Fathers’ Residential Status: Random and Fixed Effects Models

	Random effects							Fixed effects
	Resident			Nonresident				Resident			Nonresident
Variable	β		SE	β		SE	Diff. by res. status	β		SE	β		SE	Diff. by res. status
A. Spent 1+ hours (mother’s report, range: 1–5)
Annual earnings ($, in 2008)a	−.0023		.002	.0364	***	.010	***	−.0011		.002	.0282		.015	**
Average weekly hours worked (10 hours)	−.0149	***	.005	.0182		.010	**	−.0150	*	.006	.0029		.014	*
Annualized child support payment ($; in 2008)b				.0055	***	.001					.0050	***	.001
B. Engagement (mother’s report, range: 0–7)
Annual earnings ($, in 2008)a	−.0316	***	.007	.0093		.012	**	−.0238	**	.007	−.0035		.018	*
Average weekly hours worked (10 hours)	−.0648	***	.011	−.0204		.012	**	−.0517	***	.014	−.0401	*	.018
Annualized child support payment ($, in 2008)b				.0038	***	.001					.0056	***	.002
C. Responsibility (mother’s report, range: 1–4)
Annual earnings ($, in 2008)b	−.0060	**	.002	.0129		.007	*	−.0039		.002	.0146		.010
Average weekly hours worked (10 hours)	−.0169	***	.004	.0061		.007	**	−.0136	**	.005	.0059		.009
Annualized child support payment ($, in 2008)b				.0038	***	.001					.0041	***	.001
D. Father involvement composite (mother’s report)
Annual earnings ($, in 2008)a	−.0096	***	.002	.0119		.007	**	−.0075	**	.002	.0072		.009	*
Average weekly hours worked (10 hours)	−.0231	***	.004	−.0005		.006	**	−.0190	***	.004	−.0071		.008
Annualized child support payment ($, in 2008)b				.0031	***	.001					.0035	***	.001
E. Days spent with child past month (mother’s report, range: 0–30)
Annual earnings ($, in 2008)a				.2130	**	.075					.3046	**	.107
Average weekly hours worked (10 hours)				.0625		.074					.0622		.101
Annualized child support payment ($, in 2008)b				.0296	***	.006					.0350	***	.009
Number of observations	7,167			3,888				7,167			3,888
Number of unique cases	2,854			1,989				2,854			1,989

Note. Each cell represents a separate regression model and includes all time-constant and time-varying covariates in Tables 1 and 2. Diff. = difference. Res. = residential.

^aIn $l,000s.

^bIn $100s.

^*p<.05.

^**p<.01.

^***p<.001.

Annual earnings are negatively associated (but not statistically significantly so) with the resident father spending one or more hours with the child. (Because the vast majority of resident fathers spend one or more hours with the child every day, we reran this analysis using a logit model with a dichotomous variable; we found that higher earnings were negatively associated [p < .05] with the father spending time with the child every day.) Average weekly hours worked are significantly and negatively associated with spending time with the child. For each additional 10 hours worked per week, a father spent 0.0149 fewer units of time with the child—a difference that is significant but modest (about 3% of a standard deviation). For paternal engagement in activities (see Panel B of Table 3), resident fathers with higher levels of economic investment are significantly less likely to engage in activities with the focal child, although the magnitude of the relationships remains modest. Each additional $10,000 of annual earnings is linked with a lower score (−0.0316) on the 0-to-7 measure of engagement in father–child activities (2% of a mean standard deviation). Average weekly hours worked is associated with lower engagement (−0.0648, about 4% of a mean standard deviation). Furthermore, the relationship between resident fathers’ economic capacities and the 1-to-4 measure of shared responsibility (Panel C of Table 3) is negative and significant across both economic indicators, although the magnitudes of the associations are also very small (1%–3% of a standard deviation). Similarly, the composite involvement measure (in standard deviation units) that combines all the father involvement measures (time, engagement, and responsibility) into a single measure is consistently, negatively related to both economic measures. Taken together, these results suggest that for resident fathers, greater investment in the labor market is associated with significantly lower—though very modest in size—levels of paternal involvement; thus, for resident fathers, investments in breadwinning and in more direct involvement with the child appear to operate as substitutes.

For nonresident fathers, the pattern of results differs in two ways: father’s economic capacities are mostly positively related with involvement, and the strength of the association varies by the particular economic characteristic and type of involvement being considered. Higher levels of earnings (0.0364) and child support payments (0.0055) are significantly and positively associated with the father being more likely to spend one or more hours with the child in the past week—although again of very modest size (2% or less of a mean standard deviation). For each additional $100 that the father paid in total child support in a year, he engaged 0.004 more days in father–child activities. Nonresident fathers who paid more in child support also had higher levels of shared responsibility for child rearing and higher scores on the composite measure of father involvement. With respect to the number of days nonresident fathers saw their child in the past month, both annual earnings and child support payments are both positively and significantly related. All associations were miniscule to modest in size.

We also tested whether the estimates for resident versus nonresident fathers were significantly different from one another. For all eight possible comparisons (two independent variables × four dependent variables) in the random effects models, we found that all estimates significantly differ between resident and nonresident fathers (p < .05). Because all estimates for resident fathers are in a negative direction and all statistically significant estimates for nonresident fathers are in a positive direction, the significant difference points to the fact that economic capacities play a fundamentally different role in facilitating the time involvement of fathers depending on whether they live with or away from their children.

Fixed effects estimates of the relationship between father’s economic characteristics and involvement—which rely only on within-father change and hence are more conservative with respect to causal inference—are also presented in Table 3. However, this approach eliminates the between-father comparison that is of overall interest in this research. For resident fathers, the results for all measures remained negative and significant, as with the random effects (though again, the magnitudes were very small), except that earnings as linked to responsibility is no longer statistically significant. These results provide greater evidence that fathers’ investments of money and time appear to operate as substitutes when fathers live with their child(ren): As the same resident fathers increase their earnings and time in the labor market over time, they also decrease their engagement and overall involvement with the focal child.

For nonresident fathers, some—but not all—of the fixed effects estimates are also statistically significant, suggesting that the random effects results were not entirely due to differences in characteristics between fathers. Increases in the amount of child support paid by nonresident fathers are positively linked with greater involvement across all five measures. On average, nonresident fathers who increased their earnings also increased the number of days they had seen the child in the past month. These results suggest that as nonresident fathers’ financial contributions to children increase, their involvement also increases over time. The single exception to this pattern is the link between average weekly hours worked per week; there is a significant negative relationship between increasing hours worked and father–child engagement.

Again, evaluating whether these point estimates significantly differed between resident and nonresident fathers, we found that four of the eight possible comparisons were statistically significant at p < .05 or below; these all occurred where the direction of the point estimate was negative for resident fathers but positive for nonresident fathers. It is not surprising that there are fewer differences once we analyze only within-father change, yet where we did find persisting differences they support our contention that economic resources may play a different role as linked to direct involvement for resident versus nonresident fathers.

As a sensitivity analysis, we repeated our main analyses using fathers’ reports of involvement (where available, results not shown); we found no major differences in the overall relationship between fathers’ economic characteristics and time involvement, and the coefficient magnitudes were similar. Furthermore, to determine the extent to which our results are driven by fathers changing residence categories over time (mostly as coresident couples broke up), we conducted a second set of analyses (not shown) limited to fathers who were stably resident or stably nonresident, that is, those who lived with the focal child over all of Years 1–9 versus those who never lived with the child. The results were remarkably similar to our main results, suggesting that the substantive conclusions are not driven by the changing composition of the groups, as fathers (primarily) move out of the child’s household. Finally, we also conducted a cluster analysis to evaluate whether our small but differing correlations of money and time for resident versus nonresident fathers were driven by within-group heterogeneity. Using all five money and time variables, cluster analyses yielded three groups of fathers: nonresident fathers, higher-earning resident fathers and lower-earning resident fathers. We then replicated our random and fixed effects analyses separately by cluster, and the results were very similar to our main results (and the results for the two groups of resident fathers did not differ).

Discussion

In this article we have presented new descriptive information about the levels and interrelationships of fathers’ economic capacities and time involvement with children among a contemporary cohort of urban fathers over child ages 1–9. As with prior research, we found that resident fathers had much higher economic capacities—and displayed much higher levels of spending time with children, engaging in father–child activities, and sharing responsibility for coordinating children’s care—than nonresident fathers (Amato, 1998; Carlson & Berger, 2013). Taken together, our results suggest that economic capacities and contributions (money) operate quite differently as related to direct involvement (time) for resident versus nonresident fathers. For resident fathers, there appears to be a trade-off between investments in market work and time spent in more direct aspects of involvement and care: They operate as substitutes. At higher levels of work effort and earnings, resident fathers displayed significantly lower levels of time spent with children, engagement in activities, and shared responsibility. As such, fathers’ investments in financial provision appear to “count” as investments in children as collective goods, consistent with the notion of the “package deal” whereby marriage, fathering, and employment go together (Furstenberg & Cherlin, 1991; Townsend, 2002).

For nonresident fathers, across all the pairwise combinations of economic resource measures and direct involvement measures, there was only one statistically significant estimate suggesting that financial investment and time involvement appeared to operate as substitutes (hours worked and father–child engagement in the fixed effects models)—even though greater time in the labor market necessarily means that fathers have less available time for other activities, including parenting. Other than this single exception, our statistically significant regression estimates all suggested that greater economic capacity and contributions by nonresident fathers instead “go together” with being involved in other ways with their children (spending time, engaging in activities, and sharing responsibility). Therefore, these two domains of involvement (money and time) appear to operate as complements for nonresident fathers as such men do “double duty” to remain involved with their children.

Taken together, these findings extend our understanding about how fathers contribute to children’s lives and how their investments vary—in both level and implications—as a function of whether they are living with or living away from their children (and the mothers of their children). It is important to note that we cannot identify the direction of this association here, although prior research that has explicitly considered the directionality of money and time investments for nonresident fathers suggests that child support payments enhance father contact more than contact enhances payments (Nepomnyaschy, 2007). Thus, our research further suggests that, for nonresident fathers, higher levels of economic resources may be an important prerequisite for fathers to also have direct involvement with their children. Also, we have not considered differences in the association as children age—a useful topic for future research.

From a broader perspective, our results suggest that differences across fathers in what they invest in children can contribute to growing inequalities in the next generation. Men who share parental investments with biological mothers in the same household not only have higher levels of both economic resources and parental time involvement but also benefit from the economies of scale that enhance parents’ ability to jointly (and efficiently) allocate the market work and household investments they make in their children as a collective good. As such, men living with children and mothers have greater flexibility in how they enact the father role, and they also have more support in doing so. By contrast, nonresident fathers—who have lower levels of economic resources—face higher transaction costs in order to spend time and engage with their children at the same time that they have less flexibility in how they invest as parents.

Although we believe our research provides a useful extension to the literature, we also acknowledge several limitations. First, our results are relevant only to families of children born in cities of population 200,000 or more. Second, fathers are typically underrepresented in national surveys (Nelson, 2004), and the Fragile Families study also did not interview some fathers and lost some fathers to attrition (typically the more disadvantaged fathers); our results thus cannot estimate the associations of interest for the missing group of nonresident fathers. Third, we recognize that fathers may be involved in other ways that we are not measuring here (e.g., telephone calls, cards/letters, or e-mail). Fourth, we used mothers’ reports in our main analyses to avoid using the same reporter for both the independent and dependent variables, but mothers may not have accurate information about fathers’ involvement, especially for nonresident fathers (Coley & Morris, 2002). Our mostly similar findings when using measures based on fathers’ reports were thus reassuring. Fifth, although our methods partially accounted for unobserved heterogeneity, our results could still be biased by time-varying unobserved differences.

In sum, this article has provided new descriptive information about fathers’ involvement in children’s lives across residential contexts. We found important differences in how the father role is enacted for resident versus nonresident fathers. Although resident fathers appear to experience a trade-off between their time in the labor market and their time directly involved with children, nonresident fathers’ contributions appear to “go together” such that financial capabilities and contributions serve to increase other aspects of involvement. Given the low economic resources of many nonresident fathers, this circumstance may create challenges for fathers to remain actively involved in their children’s lives with respect to both money and time, with long-term implications for children.