Paternal incarceration and child-reported behavioral functioning at age 9

Social Science Research 52 (2015) 18–33

Contents lists available at ScienceDirect

Social Science Research journal homepage: www.elsevier.com/locate/ssresearch

Paternal incarceration and child-reported behavioral functioning at age 9 Anna R. Haskins ⇑ Cornell University, Department of Sociology, USA

a r t i c l e

i n f o

Article history: Received 3 June 2014 Revised 5 November 2014 Accepted 6 January 2015 Available online 21 January 2015 Keywords: Paternal incarceration Non-cognitive development Middle childhood Child self-reports Propensity Score Matching Socio-emotional behavior

a b s t r a c t Within the last few decades our understanding of the importance of non-cognitive skills for socioeconomic success has grown along with our knowledge of the deleterious impacts of paternal incarceration for child wellbeing. Given the importance of early skills and that elementary-aged children constitute the majority of children with incarcerated parents, understanding the connection between paternal incarceration and the socio-emotional component of children’s non-cognitive development is pressing. Using matching models, data from the newest wave of the Fragile Families and Child Wellbeing Study, and exploring a larger range of behavioral skills than previous literature, this paper provides estimates of the impact of paternal incarceration on children’s behavioral functioning at age 9 using children’s own self-reports. Comparisons to oft-used parent reports are made and heterogeneity by gender is explored. Findings suggest the incarceration of a father increases the antisocial behaviors children self-report, but has null effects on prosocial skill development. Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction There is converging cross-disciplinary evidence of the importance of non-cognitive skills for later education, employment and earnings outcomes—in short, one’s socioeconomic success. Somewhat independently, this work has established that a variety of non-cognitive traits such as one’s social and emotional skills, self-discipline or behavioral functioning are: consistent predictors of school success (e.g., Bowles and Gintis, 1979; Farkas et al., 1990), equally as important as cognitive skills or schooling for later-life earnings, educational attainment and labor market success (e.g., Duckworth and Seligman, 2005; Dunifon and Duncan, 1998; Heckman and Rubinstein, 2001; Heckman et al., 2006; Jencks, 1979), and dramatically shaped by early life experiences creating lasting implications for subsequent skill development (e.g., Knudsen et al., 2006). Alongside this, is mounting evidence of the deleterious ripple effects of mass incarceration—those that extend beyond the imprisoned individual to impact families, communities and other social institutions. Given its high cumulative risk (Wildeman, 2009), of particular importance are studies showing the negative impacts of a father’s incarceration for child and adolescent mental health and socio-emotional development. Moreover, paternal imprisonment, somewhat more consistently than a mother’s incarceration (see for example, Wildeman and Turney, 2014 or National Research Council, 2014 for overviews), has been found to increase aggression, depression, anxiety, attention problems, and delinquency in young children and adolescents (e.g., Geller et al., 2012; Haskins, 2014; Murray and Farrington, 2008; Murray et al., 2012b; Roettger

⇑ Corresponding author at: Cornell University, Department of Sociology, 354 Uris Hall, Ithaca, NY 14853, USA. E-mail address: [email protected] http://dx.doi.org/10.1016/j.ssresearch.2015.01.001 0049-089X/Ó 2015 Elsevier Inc. All rights reserved.

A.R. Haskins / Social Science Research 52 (2015) 18–33

19

and Swisher, 2011; Wilbur et al., 2007; Wildeman, 2010), suggesting the potential for mass paternal incarceration to present a significant impediment to a child’s healthy non-cognitive skill development and thus their future socioeconomic success. With children of the incarcerated constituting nearly 10% of the total U.S. population under the age of 18 (Travis et al., 2005), this paper builds on these strands of work by bringing new data to bear, from the latest wave of the Fragile Families and Child Wellbeing Study, on our understanding of the effect that paternal incarceration has on an important dimension of non-cognitive skill development: children’s self-reported socio-emotional skills and behavioral functioning at age 9. Special care is given to issues of selection by utilizing matching methods, accounting for a robust set of covariates, and attending to the temporal ordering of all measures included in the analyses. Moreover, examinations of whether effects vary by gender and how similar findings are to parent reports of child socio-emotional behaviors are conducted. In all, this study provides a novel extension by bringing new quantitative data into the conversation on the effects of paternal incarceration for children’s development, helping to paint a more nuanced picture of the collateral consequences of mass incarceration for child wellbeing into middle childhood. 1.1. Non-cognitive skills and the lasting importance of early skill development The multi-disciplinary interest in, and broad nature of, non-cognitive skills makes finding a common terminology difficult. Children’s non-cognitive skills can encompass dimensions of physical health or motor functioning as well as social and emotional behaviors, personality traits, or abilities linked to self-discipline and effortful control. The specific dimensions of non-cognitive skills that this paper explores are the attention, social and behavioral components of learning, which align with a child’s ability to concentrate, stay on task, cooperate, interact appropriately with peers, and exercise emotional selfregulation (for a review, see Farkas, 2003). Non-cognitive skill development is cumulative, begins during the earliest years of life, and is powerfully shaped—both negatively and positively—by experiences and environments in early childhood (Knudsen et al., 2006; Shonkoff and Phillips, 2000). During early childhood (approximately birth to age 5) the foundation for one’s skill capacities is laid, while in middle childhood (approximately ages 5–10) these skills crystalize, establishing a trajectory for future development (Blair, 2002; Kowaleski-Jones and Duncan, 1999). Thus, negative experiences—whether social, environmental or physical—occurring during the first ten years of a child’s life have the potential to influence a range of later outcomes (Duncan et al., 2007; Huston and Ripke, 2006), such as schooling, employment and earnings, all of which are linked to broader stratification processes. 1.2. Parental incarceration and child socio-emotional development and behavioral functioning The incarceration of a parent has the potential to present a number of risks to a child’s healthy emotional, behavioral and social development. When children are left to make sense of and deal with the absence of a parent, often without explanation or understanding, parental incarceration can lead to feelings of worry, confusion, loneliness, ambiguous loss, anger, depression, sleep problems, or even developmental regressions (Bocknek et al., 2009; Poehlmann, 2005). The vast majority of work looking at how a parent’s incarceration affects their offspring emphasizes impacts along the socio-emotional, delinquency and mental health dimensions of development (for a recent review, see Murray et al., 2012a). For example, among adolescents with incarcerated mothers, findings emphasize juvenile justice involvement (Cho, 2010; Shlafer et al., 2012), aggression and bullying (Myers et al., 2013) and increased depressive symptoms (Foster and Hagan, 2013; Lee et al., 2013). Likewise, paternal incarceration for adolescents carries increased probabilities of delinquency (e.g. Roettger and Swisher, 2011), alongside stress, anxiety and other antisocial behaviors (e.g. Foster and Hagan, 2013; Murray and Farrington, 2008). However, of arguably more importance are studies exploring effects of parental incarceration on younger children’s socio-emotional and behavioral functioning, given the importance of childhood for setting the foundation for healthy skill development. Studies on consequences of parental incarceration for children during early childhood show negative impacts across a range of non-cognitive skills, including physical aggression, behavior problems, and attentional capacities (Craigie, 2011; Geller et al., 2012; Haskins, 2014; Johnson, 2009; Wildeman, 2010). Effects across this pre-school age range are most robust for externalizing behaviors (aggression, acting out) and are mainly reported to be concentrated among boys. Slightly less consistent findings are reported for children who experience a parent’s incarceration during the elementary school years (i.e. middle childhood, ages 5–10). For example, work by Wilbur et al. (2007) and Wakefield and Wildeman (2011) provide evidence for increases in child externalizing behaviors and depressive symptoms after parental incarceration, while Johnson (2009) finds only marginal effects for internalizing and null effects for externalizing behaviors for children in middle childhood. Focusing mainly on antisocial behaviors, fewer studies of impacts of parental incarceration on children in middle childhood have looked at measures of prosocial non-cognitive skills, such as task completion or self-discipline, which are quite important to future socioeconomic success (Duckworth and Seligman, 2005; Heckman et al., 2006). In one recent exception, Dallaire and Zeman (2013), explore the relationship between elementary aged children’s empathic behavior, peer-reported aggression and exposure to parental incarceration. Looking at differences in reports of child empathy and aggressive behaviors across multiple groups of children experiencing a variety of forms of parental separation, they find that children experiencing current parental incarceration are, on average, rated as less empathic and more aggressive than peers in the other comparison groups.

20

A.R. Haskins / Social Science Research 52 (2015) 18–33

Additionally, while qualitative interviews have highlighted impacts of parental incarceration from the child’s perspective (e.g., Bocknek et al., 2009; Nesmith and Ruhland, 2008), to date, most large quantitative datasets employed to investigate effects have relied solely on adult reports (parent, caregiver or teacher) of child outcomes drawing critiques from some scholars in the field (Johnson and Easterling, 2012). Adult perspectives on child socio-emotional skill capacities are useful, but also only provide part of the picture. In one of the few studies that has recently explored the impact of parental incarceration on children’s behaviors using measures from multiple respondents, Dallaire and Zeman (2013) find significant differences between parent-reports of child behavior across the groups studied—with parent reports indicating children of currently incarcerated parents exhibit less empathic behaviors than the other comparison groups—while no significant differences across groups surface using outside observer and child self-reports of empathy.1 Thus, understanding the impacts of a parent’s incarceration using children’s own self-reported behaviors can add depth to our understanding of the collateral consequences of incarceration and allow for a more nuanced picture to be formed around how children themselves experience the incarceration of a father. 1.2.1. Paternal incarceration and gender differences in non-cognitive skill development By far, men are incarcerated at higher rates than women, leading to fathers accounting for 91% of all incarcerated parents (Maruschak et al., 2010). Work on fatherhood has shown men make important contributions to the development of both sons and daughters (Lamb, 2010), however, some scholarship suggests relationships with fathers are potentially more salient for sons (for a review, see Allen and Daly, 2007). Additionally, a growing body of literature shows boys appear to be more sensitive to family disruption and instability, leading to increased incidence of emotional, behavioral, social and developmental problems (Buchmann et al., 2008; Cooper et al., 2011; Foster and Hagan, 2013). Furthermore, gender differences in non-cognitive skills present in early childhood grow throughout elementary school, impacting cognitive test scores and explaining a large portion of the gender gap in later academic outcomes (DiPrete and Jennings, 2012). Along these lines, work exploring heterogeneity in effects by gender of the incarcerated parent and affected child has begun to take shape. For example, Wildeman (2010) finds paternal incarceration to increase physical aggression in young boys but not girls, increasing the risk for male intergenerational transmission of criminality. Haskins (2009, 2014) also shows that boys experiencing paternal incarceration have significantly lower non-cognitive school readiness than same-age girls, suggesting paternal incarceration plays a role in explaining the low achievement of poor and minority boys upon school entry. And, recent work by Foster and Hagan (2013) on adolescents concludes that male children may have a heightened vulnerability to both maternal and paternal imprisonment, implying that parent absence, or a break in the parent–child relationship due to incarceration, may result in a unique or extended disadvantage for boys. However, effects of a father’s incarceration may impact both boys and girls with the potential for similar or differing consequences depending on the age of the child. While there has been less recent evidence of impacts of paternal incarceration for pre-school age girls’ development, some work on children during middle childhood finds negative impacts of a father’s incarceration for both boys’ and girls’ cognitive outcomes (Haskins, 2013) and their likelihood of early grade retention (Turney and Haskins, 2014). Moreover, Fritsch and Burkhead (1981) report distinct differences between the gender of the parent incarcerated and reported child behaviors, concluding paternal incarceration to be more strongly associated with ‘‘acting-out’’ or externalizing behaviors among the school-aged children in their sample. Lastly, among adolescents, Lee et al. (2013), find paternal incarceration to increase mental health problems (e.g. depression, anxiety and PTSD) in children while maternal incarceration was more associated with increased odds of physical health problems. 1.3. Usefulness of comparisons of child outcomes across multiple respondents One of the long standing critiques of research on effects of parental incarceration has been the overreliance on parental or adult reports of child outcomes (for a review, see Johnson and Easterling, 2012). As suggested above, child self-reports of non-cognitive skill capacities can provide additional (and potentially differing, see for example Dallaire and Zeman (2013)) evidence of the impacts of paternal incarceration. Moreover, the inclusion of adult reports alongside child selfreports of similar socio-emotional and behavioral outcomes can begin to help flesh out similarities and differences in the ways respondents experience or perceive the impact a father’s incarceration has on behavioral dimensions of children’s non-cognitive skill development into middle childhood. 1.4. Study contributions In summary, this study strengthens and adds to previous work in five ways. First, it uses newly available longitudinal data from a large, contemporary and representative dataset to explore effects of paternal incarceration on the self-reported socioemotional and behavioral capacities of elementary-aged children. Second, it investigates a larger range of non-cognitive outcomes than previous research, incorporating both early delinquency and prosocial behaviors alongside the more oft-used measures of internalizing and externalizing behaviors. Third, sensitivity to effect heterogeneity by gender is explored given 1 However, in moderation analyses conduced with the group of children with currently incarceration parents, child-self reports (and not parent or outside observer reports) of empathy significantly predicted children’s peer-reported aggressive behaviors. Dallaire and Zeman (2013) suggest that higher self-reported empathy in children with currently incarcerated parents serves as a protective factor against developing aggressive peer relations.

A.R. Haskins / Social Science Research 52 (2015) 18–33

21

some of the patterns emerging from recent work. Fourth, it compares parent-reports of child behavioral skills, alongside child self-reports to bring depth to our understanding of the collateral consequences of paternal incarceration. Lastly, care is taken to address a number of methodological limitations that have plagued earlier work on the effects of parental incarceration on children. Specifically, by analyzing the rich paternal incarceration data in the Fragile Families and Child Wellbeing Study, this study first asks whether paternal incarceration is detrimentally associated with four measures of children’s self-reported socioemotional and behavioral skills at age 9. Following this, it investigates whether effects vary by child gender, and then concludes with a comparison of how different findings are based on the use of child self-reports versus parent-reports of two behavioral skill measures. To disentangle the unique impact of paternal incarceration from the effects of preexisting disadvantage, special attention is paid to the temporal ordering of all variables in the models, controls for a wide range of covariates measured prior to incarceration are also included, addressing with greater sensitivity the process of social selection, and propensity score matching techniques are employed in order to make appropriate non-cognitive skill comparisons between children with differing paternal incarceration experiences but similar economic, demographic, household, neighborhood, and parental behavior characteristics. Results highlight the usefulness of having both parent- and child-reports of behavioral capacities in understanding how the experience of paternal incarceration is perceived and suggest paternal incarceration is positively associated with the development of antisocial behaviors but is not harmful to children’s prosocial skill development. 2. Methods 2.1. Data The Fragile Families and Child Wellbeing Study (FFS) is a longitudinal birth-cohort study that follows 4898 focal children and their parents. Collected from 20 large U.S. cities between the years of 1998–2000, marital and non-marital births were randomly sampled within hospitals that were stratified by labor market conditions and policy environments (for a complete description of the sample and design see Reichman et al., 2001).2 For mothers, baseline interviews took place in hospitals within 48 h after the birth of the focal child and for fathers, soon thereafter. Since the baseline wave, four additional followup waves of phone interviews have occurred, taking place when the child is approximately 1, 3, 5, and 9 years old. Each wave includes separate interviews of each parent, in-home direct assessments of the child and their home environment (starting at the year 3 follow-up), and for the most recent wave—when the child is 9 years old and has entered elementary school—it includes a teacher survey, a focal child survey, and administrative data on the child’s elementary school. The baseline response rate for the nationally representative sample of mothers is 86%, while for fathers it is slightly lower at 79%. Follow-up interview response rates for both parents across waves can be found in the online supplement. The FFS is one of the few broadly representative data sources currently available to explore contemporary questions related to paternal incarceration’s impacts on child outcomes. Not only does it follow both parents over time as their child grows,3 but also has great response rates given the focal population, and sizeable variation by paternal incarceration experiences to explore effects. Moreover, the restricted data, contain a wealth of background, demographic, environmental, household, health, neighborhood and economic information to include as controls in analyses. Analyses for this paper take advantage of information from across the five current waves of data, covering the first 9 years of the focal child’s life. The overall analytic sample includes all children in the FFS, however anyone missing information on the socio-emotional and behavioral outcomes explored or with paternal incarceration experiences occurring prior to age 1 are dropped, giving, on average, final samples of N = 2150 (for child-reported outcomes) and N = 2015 (for parent-reported outcomes). Subgroup analyses by gender split these analytic samples to, on average, n = 1127 for boys and n = 1023 for girls for the child-reported outcomes. For the parent-reported outcomes, subsamples are n = 1066 for boys and n = 949 for girls. With these sample sizes, the calculated minimum detectable effect size (MDES)4 for the overall child-reported outcomes is 0.139, while the MDES for the overall parent-reported outcomes is slightly higher at 0.143. Given the smaller gender subgroups, the MDES for these subsamples range from 0.193–0.213. Table 1 provides a descriptive snapshot of the overall analytical sample by paternal incarceration status, providing individual sample counts for each child and parent outcome investigated. Missing covariates are preserved with five multiply imputed data sets derived in Stata 12 using the ICE (Imputation by Chained Equations) procedure.5 Statistical analysis is then done on each individual dataset and averaged to yield a final single set of results (Rubin, 1987; Royston, 2005a, 2005b).

2 The FFS data, when weighted, are nationally representative of both marital and non-marital births to parents residing in cities with populations of 200,000 or more. However, non-marital births were oversampled, thus nearly three-quarters of the parents in the study are unmarried (n = 3712) at the baseline wave, while 1186 are married. 3 An attribute of the study that helps to validate the reliability of the child and parent measures in addition to providing information (via maternal, and recently child self-reports) about disadvantaged fathers otherwise unavailable given their frequent underrepresentation in surveys (Hernandez and Brandon, 2002; Reichman et al., 2001). 4 Calculated using the PowerUp! tool created by Dong and Maynard (2013). 5 Differences in descriptive characteristics across the imputed and non-imputed datasets for the covariates included in the analytic models were negligible.

Variable names

dir

Father never incarcerated (G1)

22

Table 1 Weighted means and standard deviations for dependent and independent variables by paternal incarceration status. Father incarcerated Btw YR1 and YR9 (G2)

G1 vs. G2 t-test sig. Ns

SD

Mean

SD

0.26 0.3 0.21 0.04

(0.95) (0.93) (1.23) (1.02)

0.2 0.14 0.05 0.03

(1.14) (1.07) (1.17) (1.03)

⁄⁄⁄

0.25 0.16

(0.76) (0.81)

0.09 0.2

(1.49) (1.56)

⁄⁄⁄

Maternal demographic and household characteristics child age at YR9 (in months) child grade at YR9 child race Black child race White child race Hispanic child gender Male child Low Birth Weight child Healthy Maternal cognitive (0–15) Maternal Self-Control (6–24) Maternal Age at 1st Birth (13–45) Maternal Education (1–4) Mother Cohabiting with Father Mother Married to Father Maternal Parenting Stress (0–12) Maternal Anxiety Maternal Depression # of Maternal Bio Kids (1–16) Grandparent in HH # of Children in HH (0–8)

111.61 3.13 0.31 0.4 0.29 0.6 0.07 0.99 7.29 18.17 24.72 2.59 0.18 0.64 4.7 0.03 0.13 2.06 0.14 1.07

(3.55) (0.69)

112.1 3.1 0.53 0.1 0.38 0.55 0.07 0.97 6.81 17.81 21.2 1.91 0.25 0.4 4.81 0.03 0.15 2.26 0.23 1.43

(3.96) (0.68)

NS NS

(1.36)

⁄⁄

Paternal Demographic and Psycho-Social Characteristics Paternal Age (15–53) Father Employed Father US Citizen Paternal Cognitive (0–15) Paternal Education (1–4) Paternal Self-Control (6–24) Paternal Drug and Alcohol Problems Paternal Domestic Violence Father had Two Bio-Parent HH at 15 Father’s Bio Father Involved Paternal Multi-partner Fertility Paternal Anxiety

31.13 0.9 0.69 6.56 2.64 19.04 0.04 0.07 0.62 0.79 0.24 0.03

(7.28)

27.37 0.84 0.8 5.92 1.91 17.2 0.12 0.16 0.48 0.68 0.36 0.05

(7.38)

⁄⁄⁄

Outcome variables Child-reported non-cognitive SDQ Externalizing Problems SDQ Internalizing Problems Delinquent Behavior Task Completion Parent-reported non-cognitive CBCL externalizing problems CBCL internalizing problems

+

⁄⁄⁄ ⁄⁄⁄ ⁄⁄

⁄⁄⁄

G1

G2

Total

1603 1601 1587 1591

559 556 555 550

2162 2157 2142 2141

1517 1472

525 516

2042 1988

in PSM

(2.92) (3.44) (6.11) (1.09)

(2.62)

(1.20) (1.31)

(2.97) (1.10) (3.60)

⁄⁄⁄ ⁄⁄⁄ ⁄⁄

NS NS ⁄⁄

(2.38) (3.70) (4.78) (0.82)

+ ⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄ ⁄⁄⁄

(2.91)

(1.35)

NS NS NS NS ⁄⁄⁄

⁄⁄⁄ ⁄⁄⁄

(2.68) (0.86) (3.75)

⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄

x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x

A.R. Haskins / Social Science Research 52 (2015) 18–33

Mean

0.07 26.77

Economic Indicators Poverty Status (1–5) Child Living in Public Housing Neighborhood Unsafe

2.25 0.07 0.08

(1.33)

Census Tract Characteristics % of population White % of population Black % of female pop. of childbearing age % of HHs female-headed w/children < 18 mean # of persons per HH % of 25 + population with HS + education % of 25 + population with BA + education % of civilian labor force unemployed % of housing units vacant % of housing units renter-occupied median housing value in dollars in 1999 % of HH on public assistance % of families below poverty level in 1999 % of families w/1999 income < $10 K % of families w/1999 income $10–14,999 % of families w/1999 income $15–24,999 % of families w/1999 income $25–34,999 % of families w/1999 income $35–49,999 % of families w/1999 income $50–74,999 % of families w/1999 income $75–99,999 % of families w/1999 income $100–149,999

0.39 0.26 0.52 0.17 2.82 0.72 0.25 0.1 0.07 0.55 154,520 0.08 0.18 0.12 0.06 0.12 0.12 0.15 0.17 0.1 0.08

(0.35) (0.33) (0.07) (0.13) (0.75) (0.18) (0.22) (0.07) (0.05) (0.28) (158,760) (0.08) (0.16) (0.11) (0.05) (0.06) (0.06) (0.06) (0.07) (0.06) (0.08)

(8.38)

0.13 24.36

⁄⁄

(9.98)

⁄⁄⁄

3.17 0.09 0.19

(1.26)

⁄⁄⁄

0.26 0.43 0.52 0.23 2.76 0.68 0.15 0.13 0.09 0.57 116,626 0.1 0.23 0.16 0.07 0.14 0.14 0.16 0.18 0.08 0.05

(0.29) (0.39) (0.07) (0.14) (0.53) (0.14) (0.14) (0.09) (0.08) (0.25) (170,238) (0.12) (0.16) (0.14) (0.05) (0.06) (0.05) (0.05) (0.08) (0.05) (0.05)

Interview city (20 indicator variables) Notes: Descriptives of controls run on largest outcome sample (N = 2162); 20-cities weights used; t-tests run on unweighted data.

⁄⁄⁄ ⁄⁄⁄

⁄⁄⁄ ⁄⁄⁄

NS ⁄⁄⁄

NS ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄

NS ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄

x x x x x x x x x x x x x x x x x x x x x x x x x x x x x

A.R. Haskins / Social Science Research 52 (2015) 18–33

Paternal Depression Paternal Contact with Child in days (0–30)

23

24

A.R. Haskins / Social Science Research 52 (2015) 18–33

2.2. Measures 2.2.1. Child-reported socio-emotional and behavioral functioning A total of four child-reported non-cognitive behaviors are explored, each drawn from the child survey given during the 9 year follow-up wave. Together they encompass a broad range of socio-emotional indicators, from externalizing and internalizing problems to task completion and delinquent behaviors, and are standardized scales or adaptations of measures used in other nationally representative surveys of children. The six-item externalizing (alpha = 0.76) and eight-item internalizing (alpha = 0.78) behavior problems measures are created with items from the Self-Description Questionnaire (SDQ) (Marsh, 1990), a self-administered survey that is widely used as a standardized measure of self-concept for pre-adolescents. It is designed to measure children’s self-perceptions of their school abilities, peer relationships and problem behaviors. Children were asked to rate their frequency of emotions and behaviors on a scale from 0 to 3, with responses including ‘‘not at all true,’’ ‘‘a little bit true,’’ ‘‘mostly true’’ and ‘‘very true.’’ Scores for each scale were summed with higher numbers indicating higher externalizing or internalizing behavior problems. The task completion measure (alpha = 0.59) is comprised of child responses to five items and is drawn from the Child Development Supplement of the Panel Study of Income Dynamics. This measure taps a child’s perseverance (Furstenberg et al., 1999); children are asked to rate the frequency of which they do a number of task persistent behaviors on a scale from 0 to 3, with responses including ‘‘never,’’ ‘‘rarely,’’ ‘‘sometimes’’ and ‘‘always.’’ Scores are then summed with higher numbers indicating more positive task completion behavior. The delinquent behaviors measure (alpha = 0.70) is drawn from the ‘‘Things You Have Done’’ (TYHD) scale (MaumaryGremaud, 2000) which has been used to measure child delinquent behavior on the National Longitudinal Survey of Youth. This scale was created by summing child reported responses (yes = 1, no = 0) to a list of 17 destructive, illegal or antisocial items, with higher scores representing more self-reported delinquent behaviors. For a full list of the survey questions that make up each scale, see the online supplement. 2.2.2. Parent-reported behavioral functioning The two measures of parent-reported non-cognitive behaviors used for comparison represent child externalizing and internalizing problem behaviors. These are based on maternal/primary caregiver reports of items from the school-age version of the Child Behavior Checklist 6–18 (CBCL), one of the most widely used standardized measures in child psychology for evaluating maladaptive behavioral and emotional problems (Achenbach and Rescorla, 2001). To compute the scores for the externalizing and internalizing constructs, responses to each of the items (0 = not true; 1 = somewhat/sometimes true; 2 = very true/often true) in their respective scales were summed so that higher numbers would indicate worse behavioral problems. The CBCL externalizing problem behavior subscale (35 items; alpha = 0.91) taps parents’ perceptions of their child’s aggressive, rule-breaking and acting out behaviors, while the internalizing problem behavior subscale (33 items; alpha = 0.88) covers depression, withdrawal and anxiety behaviors; see the online supplement for the survey questions that make up these scales. 2.2.3. Paternal incarceration The key independent variable is first-time paternal incarceration between the ages of 1 and 9. Many of the fathers in the FFS experience incarceration. Fig. 1 shows the progression of paternal incarceration for children over waves. At the first follow-up wave (year 1), when the focal child is age 1 and paternal incarceration is first systematically measured, approximately 30% of the fathers in the study have experienced incarceration at some point in their lives, and this increases to nearly 46% by age 9—totaling just over 2300 dads. For this contemporary sample of urban school-aged children, paternal incarceration is not a rare life event.

70.0%

63.8%

60.0% 49.6%

50.0%

38.8%

40.0% 30.0%

49.3% 41.6%

45.4% 38.1%

29.8%

Ever Never Unknown

16.5%

20.0% 11.6%

10.0%

6.5%

9.1%

0.0% Year 1

Year 3

Year 5

Year 9

Notes: Non-imputed (unknowns included), N=4898 Fig. 1. Prevalence of paternal incarceration in Fragile Families over waves in percentages.

25

A.R. Haskins / Social Science Research 52 (2015) 18–33

60.0% 50.0%

50.2% 50.0%

40.0% 32.9% 33.3% Boy

30.0%

Girl

20.0%

16.9% 16.7%

10.0% 0.0% Never

Before or by YR1

Btw YR1 and YR9

Notes: Imputed; N=4898 Fig. 2. Exposure to paternal incarceration in FFS by year 9 and gender in percentages.

However, complicating any work in this area is the fact that incarceration does not happen at random, and many of the same factors that predict incarceration also predict child outcomes. The incarcerated are disproportionally poor, African American, and poorly educated (Western and Beckett 1999). Thus, children of the incarcerated and formerly incarcerated are likely to suffer from forms of socio-structural disadvantage independent of their parent’s incarceration. Second, controlling for demographics, fathers who become incarcerated exhibit higher levels of antisocial and deviant behavior, such as domestic violence, impulsivity, and substance abuse (Murray et al., 2012b), and these behaviors also have consequences for child outcomes. Because these differences likely impact child non-cognitive skill development outside of the father’s incarceration, it is necessary to contend with the possibility that both types of preexisting differences account for much of the disadvantage these children experience. To address this concern, the paternal incarceration measure used is constructed to attend to the temporal ordering of covariates most strongly associated with these forms of selection. Created from a combination of mother and father reports of the father’s current or previous incarceration status (ever or never) across study waves (beginning with the year 1 followup wave when the child is age 1 and ending at the year 9 follow-up),6 children’s paternal incarceration experiences at age 9 are indicated in one of two ways: (1) children with fathers who have no discernible incarceration histories or experiences (as reported by either mother or father consistently across the five waves) are indicated as having ‘‘never’’ incarcerated fathers, and serve as the comparison group in analyses, while (2) children whose fathers experienced first-time imprisonment sometime between the year 1 and year 9 follow-up interviews (and not earlier) are placed in the treatment group.7 Because no direct question was asked of FFS mothers or fathers at baseline/child’s birth about past or current episodes of incarceration, for the purposes of this paper, children with reports of ‘‘ever’’ paternal incarceration status at age 1 are dropped from the analytic sample.8 Descriptively, this group of children—children with fathers who experienced incarceration at any point before or by the child’s first birthday—are on average more disadvantaged than their ‘‘never’’ peers but appear quite similar across a number of characteristics to their ‘‘between year 1 and year 9’’ counterparts (see the online supplement for more specific descriptive comparisons across children in the FFS experiencing paternal incarceration). Recall from Fig. 1 that nearly 30% of the full sample consists of children with paternal incarceration experiences ‘‘before or by year 1’’. This sample refinement is necessary in order to attend most carefully to the temporal ordering of controls and provide as unbiased estimates of the effect of paternal incarceration as is possible. While fathers in the ‘‘between years 1 and 9’’ group account for a smaller number of the proportion of incarcerated fathers, they are more appropriate for estimating effects since their first-time incarceration occurred after the collection of relevant baseline and year 1 covariates. Fig. 2 shows the breakdown of paternal incarceration status by gender for the two groups of children in the analytic sample (‘‘never’’ and ‘‘between year 1 and year 9’’), as well as for the children dropped from analyses due to their father’s incarceration occurring prior to year 1 (‘‘before or by year 1’’).

6 At each wave mothers are asked, through a variety of interview questions, if their child’s father is currently incarcerated (at the point-in-time of the interview) or has ever spent time in jail or prison; fathers are asked if they have ever been imprisoned. If either mother or father answer yes to any question related to paternal incarceration, then the father is indicated as ‘‘ever’’ incarcerated for that and subsequent waves. 7 These fathers were indicated as ‘‘never’’ incarcerated at year/age 1 but by year/age 9 had an incarceration episode reported by either the child’s mother or by the father himself. 8 This group of children’s could have had fathers who experienced incarceration at any point before their birth up until the year 1 follow-up interview when the child is approximately one year old. This group is the largest in the data as well as the most difficult to make causal claims about because incarceration potentially occurred before the measurement of important baseline covariates. However, while this is still an important group to study, it is hard to differentiate the direction of influences which renders any estimates of the effect of paternal incarceration on outcomes for children in this group susceptible to bias.

26

A.R. Haskins / Social Science Research 52 (2015) 18–33

2.2.4. Controls The wealth of information available in the restricted FFS data allows for the control of a host of characteristics of mothers, fathers, and their children likely to be associated with paternal incarceration and children’s behavioral functioning. These include basic demographic and household characteristics, measures of health and economic wellbeing, an indicator for interview city, a number of contextual (census-tract) characteristics, and specific measures of maternal and paternal psychosocial and deviant behaviors, all measured prior to the father experiencing his first incarceration. Adjusting for this last set of indicators helps diminish concerns that parental behaviors both drive a father’s incarceration and impact child behaviors. Moreover, estimates of the effect of paternal incarceration are only plausible if included controls adequately address both socio-structural and deviant behavior selection. This requires not only nuanced measures of both, but also measures that precede incarceration (since both may be impacted by incarceration). In order to maintain appropriate time ordering between the dependent, explanatory and control variables, all controls included in the analyses are measured at either the baseline or year 1 follow-up interviews or are assumed fixed traits.9 A list of all included controls—57 in total—along with descriptive statistics by paternal incarceration status are provided in Table 1. 2.3. Analytic approach This paper utilizes propensity score matching (PSM) (Caliendo and Kopeinig, 2008; Dehejia and Wahba, 1999; Rosenbaum and Rubin, 1983) to estimate the relationship between paternal incarceration and child-reported non-cognitive behaviors at age 9 while controlling for important pre-incarceration characteristics of children and families. PSM models estimate the ‘‘treatment effect’’ of having an incarcerated father on children’s outcomes by simulating ‘‘treatment’’ and ‘‘control’’ groups from the observational FFS data. This matching technique allows for the appropriate outcome comparisons, via the use of a reference group of children who do not experience paternal incarceration, but are similarly at risk, based on the observed socioeconomic, demographic, neighborhood, health, and parental behavior covariates included in the matching model. The process of propensity score matching creates a high degree of covariate balance between the treatment and control groups, which allows for a more accurate comparison of ‘‘like with like.’’ Its main purpose is to ensure that the observed characteristics that would predict having an incarcerated father are balanced across those children who, in reality, do and do not experience paternal incarceration (Augurzky and Schmidt, 2001). PSM analyses are only conducted on the group of children who experience a first-time incarcerated father between ages 1 and 9 and their matched ‘‘never’’ controls. Analyses are restricted to cases within the region of common support, meaning if there isn’t an appropriate control match for a child who receives the treatment—paternal incarceration—they are not included in the analyses. Propensities are generated via a probit regression model predicting selection into paternal incarceration. The kernel matching technique is employed using a Gaussian kernel and a bandwidth of 0.08 to estimate the average treatment effect.10 After propensity scores are estimated,11 a test that the PSM technique achieved covariate balance is conducted. This was done using the PSTEST command in Stata 12 which uses a standard bias. Covariate balance statistics for each model are reported in their respective tables. For graphs of the distribution of common support for treatment and control groups across all outcomes see the online supplement. 3. Results 3.1. Descriptive statistics As demonstrated in Fig. 1, the FFS data display high patterns of early exposure to paternal incarceration. Table 1 presents descriptive statistics by paternal incarceration status for children in the analytic sample. Significant differences in unadjusted means across children with and without incarcerated fathers surface for 81% of the covariates included in the PSM models. While boys and girls experience paternal incarceration nearly equally (see Fig. 2), children with incarcerated fathers (for the first-time between ages 1 and 9) are more likely to be Black, experience higher levels of poverty, live in multigenerational households, and reside in neighborhoods perceived to be unsafe and with higher percentages of female headed households, unemployment, and families living below the poverty line. Parents of these children have lower levels of education and cognitive ability, are less likely to be married at the time of the child’s birth and are younger, with lower levels of self-control (more impulsive behaviors). The fathers who become incarcerated between years 1 and 9 also have more problems with drugs, alcohol, and domestic violence. Across the four child-reported and two parent-reported non-cognitive behaviors explored, significant unadjusted mean differences also surface. Children with incarcerated fathers score consistently 9 Cognitive ability is measured at the year 3 follow-up wave using the Wechsler Adult Intelligence test. It is considered to be a fixed trait of parents and therefore I feel comfortable including it as a pre-treatment control. Additionally, I make the assumption that paternal and maternal self-control (see Appendix E for scale components), measured at the year 1 follow-up, are fixed traits and treat them as pre-treatment controls. 10 Kernel matching (as opposed to nearest neighbor and radius matching which offer more conservative estimates because they do not use all the available cases) minimizes variance by using weighted averages of all cases in the control group to construct the outcome estimate (Caliendo and Kopeinig, 2008). These weights depend on the distance a control observation is from the treated cases based on the outcome being estimated. 11 Given my use of multiple imputed datasets, I produced average estimates of standard errors and the effect of the treatment on the treated (ATT) using Rubin’s procedure for combining estimates across imputed datasets (Allison, 2001).

27

A.R. Haskins / Social Science Research 52 (2015) 18–33 Table 2 Propensity score matching results for Child-Reported non-cognitive outcomes. Child self-reported non-cognitive skills

Paternal incarceration Externalizing behavior Internalizing behavior Delinquent behavior Task completion

Difference ⁄

0.183 0.187⁄ 0.153⁄ 0.073

Mean bias before/after matching

SE 0.06 0.06 0.06 0.058

T-statistic 3.036 3.096 2.539 1.257

Matched pairs

Before

After

% Reduction

N

Treated

Control

(raw)

(matched)

in bias

2162 2157 2142 2141

552 550 549 542

1605 1603 1589 1593

20.4 20.5 20.1 20.2

2.9 2.9 2.9 2.9

85.8% 85.8% 85.6% 85.6%

Notes: Kernel matching model estimates shown. See Table 1 and Methods section for a complete list of variables used in the models predicting the treatment. Analyses are unweighted and done on imputed data. Matched pairs indicate the average number of treated and control observations on common support. Bolding indicates statistical significance. Significance levels are the following: ⁄p < .05; ⁄⁄p < .01; ⁄⁄⁄p < .001 (two-sided).

and significantly worse across all outcomes, indicating higher externalizing, internalizing and delinquent antisocial behaviors and lower prosocial task completion skills. 3.2. Propensity score matching models 3.2.1. Child self-reports, overall Table 2 presents PSM results for the treatment effect of paternal incarceration on child-reported externalizing and internalizing behavior problems, task completion skills and delinquent behaviors. Beginning with the first row, children who experience first-time paternal incarceration at some point during the ages of 1 and 9, compared to their matched never counterparts, report nearly 1/5 of a standard deviation (SD) more externalizing behaviors (b = 0.183, p < 0.05). This measure represents children’s self-reports of feelings of aggression and episodes of inattention and acting out. The point estimate for internalizing problem behaviors (b = 0.187, p < 0.05), presented in the second row, also suggests that paternal incarceration increases child self-reports of feelings of depression, anxiety, shame and frustration. This effect is also significant and similar in size to that of externalizing behaviors, also equaling about 1/5 of a standard deviation. The third row of Table 2 presents differences by paternal incarceration status for child-reported delinquent behaviors. These estimates show children with incarcerated fathers report significantly higher frequencies of delinquent behaviors than similarly matched elementary-aged peers with no paternal incarceration experiences by age 9. Effect sizes for this noncognitive outcome are slightly smaller—in the 1/6 of a SD range (b = 0.153, p < 0.05)—but still statistically significant. Lastly, unlike the above three measures—externalizing, internalizing and delinquency behaviors—task completion is a prosocial non-cognitive skill and carries the assumption that paternal incarceration would decrease these behaviors in children. While the point estimate, found in row 4, is in the expected negative direction, differences between treatment and control groups for this measure are not statistically significant. After matching children based on a robust set of socio-demographic, economic, behavioral and contextual characteristics predictive of both paternal incarceration and child socio-emotional capacities, there seems to be no differences in positive task completion behaviors for children with and without incarcerated fathers. Child self-reports of non-cognitive problem behaviors during elementary school help flesh out impacts of paternal incarceration from the child’s perspective. As mentioned earlier, the development of behavioral and socio-emotional capacities during middle childhood is especially important for the building of self-confidence, healthy relationships and social competence. Moreover, studies have found that strong non-cognitive abilities in the form of positive socio-emotional and behavioral skills shape proximate and long-term educational trajectories and outcomes (Carneiro and Heckman, 2003; McLeod and Kaiser, 2004; DiPrete and Jennings, 2012) such that the results in Table 2 suggest that the incarceration of a father significantly increases many of the problem behaviors that can hinder healthy development of important non-cognitive capacities. Treatment effects of paternal incarceration were shown to be significant and detrimental across three of the four childreported outcomes—task completion being the exception. In all, effect sizes in the 1/6 to 1/5 SD range suggest that for these self-reported problem behaviors, children with incarcerated fathers potentially experience schooling setbacks within the range of one to two months compared to otherwise similarly disadvantaged children. 3.2.2. Child self-reports, gender sub-groups The four panels of Table 3 present PSM results for sub-group analyses by gender of the effect of paternal incarceration on child-reported behavioral functioning. For externalizing (b = 0.248, p = 0.05), internalizing (b = 0.259, p = 0.05), and delinquent behaviors (b = 0.281, p = 0.05)—panels A, B, and C, respectively—boys with incarcerated fathers self-report more of these antisocial behaviors than their matched same-gender peers with no paternal incarceration experiences. The magnitude of these point estimates, representing effect sizes in the 1=4 of a SD range, are quite large, suggesting boys with incarcerated fathers experience socio-emotional and behavioral schooling setbacks in the two to three month range due to difficulties with externalizing, internalizing and delinquent behaviors. However, the differences between matched girls with and without incarcerated fathers across these three non-cognitive outcomes—externalizing, internalizing, and delinquent behaviors—are not statistically significant. Effects are in the expected

28

A.R. Haskins / Social Science Research 52 (2015) 18–33

Table 3 Propensity score matching results for Child-Reported non-cognitive outcomes by gender. Mean bias before/after matching

Paternal incarceration

Difference

Panel A: Externalizing behavior Within boys 0.248⁄ Within girls 0.095

Before

After

% Reduction

SE

T-statistic

N

Treated

Control

(raw)

(matched)

in bias

0.107 0.078

2.308 1.217

1134 1028

284 253

843 762

22.2 19.1

3.7 3.1

83.3 83.8

281 251

841 762

22.3 19.2

3.6 3.1

83.8 83.8

281 251

835 754

21.6 18.9

3.4 3.1

84.2 83.6

280 247

834 758

22.0 18.9

3.3 3.2

85.0 83.1

Gender difference

Difference

SE

Confidence interval

Boys–girls

0.153

0.132

[ 0.107

0.413]

0.101 0.086

2.553 1.176

1131 1026

Panel B: Internalizing behavior Within boys 0.259⁄ Within girls 0.102 Gender difference

Difference

SE

Confidence interval

Boys–girls

0.157

0.132

[ 0.103

0.417]

0.108 0.069

2.609 1.139

1123 1019

Panel C: Delinquent behavior Within boys 0.281⁄ Within girls 0.078 Gender difference

Difference

SE

Confidence interval

Boys–girls

0.203

0.128

[ 0.048

0.097 0.082

0.866 1.199

Panel D: Task completion behavior Within boys 0.086 Within girls 0.098

Matched pairs

0.454]

1121 1020

Gender difference

Difference

SE

Confidence interval

Boys–girls

0.012

0.127

[ 0.237

0.261]

Notes: Kernel matching model estimates shown. See Table 1 and Methods section for a complete list of variables used in the models predicting the treatment. Analyses are unweighted and done on imputed data. Matched pairs indicate the average number of treated and control observations on common support. Bolding indicates statistical significance. Significance levels are the following: ⁄p < .05; ⁄⁄p < .01; ⁄⁄⁄p < .001 (two-sided).

direction (that of increasing self-reports of these problem behaviors) but are between half to a third the size of the boys. Tests of the gender difference between boys and girls are also not significant, however, given the smaller sample sizes and wide confidence intervals, estimates are likely imprecise as there is inadequate power to detect effects with point estimates under the minimal detectible effect size threshold of 0.193 SD (see Section 2.2 for MDES calculations). With regard to the prosocial skill of task completion, presented in panel D, elementary boys’ and girls’ paternal incarceration experiences do not seem to significantly impact their perceived ability to persist and stay on task. 3.2.3. Comparisons with Parent-Reported Socio-Emotional Skills Panel A of Table 4 presents PSM results for the treatment effect of paternal incarceration on the two age 9 parent-reported behavioral outcomes explored, externalizing and internalizing problem behaviors from the CBCL. Differences by paternal incarceration status are first presented for the overall FFS analytic sample and then by gender sub-group. Next, these estimates are compared to the parallel child-reported estimates originally presented in Tables 2 and 3. Higher numbers for these outcomes indicate worse parent- or child-reported behavior problems. Starting with parent-reported externalizing behavior (rows 1–3 of Panel A), it is clear that adult caregivers of children experiencing paternal incarceration for the first-time between ages 1 and 9 report much higher levels of this problem behavior. This is true for the overall FFS sample (b = 0.330, p = 0.01), as well as within the boy (b = 0.430, p = 0.01) and girl (b = 0.272, p = 0.05) subgroups. Effect sizes for the parent-reported measure of this non-cognitive outcome are nearly twice as large as the child self-reports (presented in the corresponding rows of Panel B), ranging from nearly 1/3 to just over 2/5 of a standard deviation. Moreover, comparisons across reporting source present some slightly differing conclusions by gender. Parents/caregivers of children experiencing paternal incarceration report significantly higher incidence of these aggressive, rule breaking or acting out behaviors for both boys and girls, while child self-reports appear to suggest that the overall effect of paternal incarceration is mainly driven by impacts on boys externalizing behaviors and not girls. Tests of the difference between parent- and child-reports of externalizing behaviors are not statistically significant, however the likelihood of imprecision is high given the wide confidence intervals presented in Panel C. Comparisons of internalizing behavior (rows 4–6 of Panel A) present interesting and somewhat different patterns. The statistically significant point estimate of 0.174 (p = 0.05) for the overall sample suggests that paternal incarceration increases parent-reports of displays of depression, withdrawal or anxiety related behavior in their children. Unlike with externalizing

29

A.R. Haskins / Social Science Research 52 (2015) 18–33 Table 4 Propensity score matching results comparing Parent-Reported versus Child-Reported externalizing and internalizing behaviors. Panel A Parent reports (CBCL)

Panel B Child reports (SDQ)

Panel C Reporting difference

Matched pairs Paternal incarceration

Diff

Matched pairs

Parent–child

SE

T-stat

N

t

c

Diff

SE

T-stat

N

t

c

Diff

SE

Confidence interval

Externalizing behaviors Overall 0.330⁄⁄ Within boys 0.430⁄⁄ Within girls 0.272⁄

0.078 0.102 0.113

4.239 4.229 2.402

2042 1079 963

519 271 230

1518 799 721

0.183⁄ 0.248⁄ 0.095

0.060 0.107 0.078

3.036 2.308 1.217

2162 1134 1028

552 284 253

1605 843 762

0.147 0.182 0.177

0.098 0.148 0.137

[ 0.046 [ 0.108 [ 0.092

0.339] 0.472] 0.446]

Internalizing behaviors Overall Within boys Within girls

0.074 0.111 0.114

2.348 1.591 2.00

1988 1054 934

509 258 227

1456 780 696

0.187⁄ 0.259⁄ 0.102

0.060 0.101 0.086

3.096 2.553 1.176

2157 1131 1026

550 281 251

1603 841 762

0.013 0.082 0.127

0.095 0.150 0.143

[ 0.199 [ 0.376 [ 0.153

0.173] 0.212] 0.407]

0.174⁄ 0.177 0.229⁄

Notes: Kernel matching model estimates shown. See Table 1 and Methods section for a complete list of variables used in the models predicting the treatment. Analyses are unweighted and done on imputed data. See Tables 2 and 3 for balance statistics for Child-Reports; see Appendix B for balance statistics for Parent-Reports. Bolding indicates statistical significance. Significance levels are the following: ⁄p < .05; ⁄⁄p < .01; ⁄⁄⁄p < .001 (two-sided).

behavior, point estimates across the overall parent- and child-reports for internalizing behavior are quite similar in magnitude (0.174 v. 0.187). While effect sizes for this outcome are smaller than those for externalizing behaviors, they remain meaningful, ranging from just under 1/5 to nearly 1/4 of a standard deviation unit. Even though overall point estimates are similar across reporters, interesting differences by gender surface. For parent-reports (Panel A), it appears girls’ (and not boys’) internalizing behaviors drive the observed overall effect. However, comparisons with child self-reports (Panel B) show the overall impact could instead be driven by differences among boys (and not girls). Without larger samples, it is difficult to conclude with certainty if there are true differences between these parent- and child-reported non-cognitive behaviors. Tests of the difference between parent and child show no statistically significant differences, however, the width of the confidence intervals around the different point estimates indicate that there may be imprecision in the estimates due to inadequate statistical power.

4. Discussion The socio-emotional and behavioral dimensions of non-cognitive development are cumulative and dramatically shaped by early life experiences—such as paternal incarceration—creating lasting implications for children’s subsequent development and wellbeing (Knudsen et al., 2006; Shonkoff and Phillips, 2000) with the potential to influence a range of later outcomes such as schooling, employment and earnings. Drawing from work demonstrating the importance of non-cognitive skills for later socioeconomic success, this study contributes to an increasingly important body of literature that focuses on understanding the implications of mass incarceration for inequality among contemporary cohorts of American children. It considers how paternal incarceration may affect children’s behavioral functioning and socio-emotional skill development by age 9, relying for the first time on children’s self-reports of pro- and anti-social behaviors. Using matching methods and attending to selection concerns, findings suggest that experiencing first-time paternal incarceration between the ages of 1 and 9 is associated with higher child-reported anti-social behaviors, including internalizing, externalizing and early delinquency problems. Propensity score models indicate the magnitude of the overall effect of paternal incarceration on these antisocial behaviors is not inconsequential, suggesting a schooling setback in the range of one to two months. However, no detrimental impacts of paternal incarceration are found for a measure of children’s prosocial skills—task completion—suggesting that there may be types of non-cognitive skills that paternal incarceration has less of an impact on. While promising, this finding is far from conclusive, as there stands potential for measurement concerns. While task completion is the only prosocial measure examined and the only behavioral measure with non-significant differences, it is also the measure with the lowest internal reliability (alpha = 0.59), leaving room for potential measurement concerns to play a role in explaining this differential finding. Nevertheless, very few studies to date (see Dallaire and Zeman, 2013; Myers et al., 2013 for exceptions) have explored the impact of parental incarceration on children’s prosocial skill development and hopefully these findings will stimulate more work in this area. Prosocial skills are quite important to future socioeconomic success (Duckworth and Seligman, 2005; Heckman et al., 2006) and evidence of null effects of paternal incarceration on prosocial skills is promising and may lead to better targeted policy interventions as we fine-tune our understanding of the ways paternal incarceration is most detrimental to children’s development. Just as previous work has documented the deleterious effects of paternal incarceration for parent-reports of pre-school age boys’ behavior (e.g. Haskins, 2014; Wildeman, 2010), analyses by gender subgroup across this diverse set of childreported non-cognitive outcomes demonstrates that among nine-year-old boys in the FFS sample the negative impacts of paternal incarceration persist into middle childhood. Among girls, associations are in the expected direction—increasing

30

A.R. Haskins / Social Science Research 52 (2015) 18–33

self-reports of antisocial behaviors—but the magnitude of the effect is much weaker and does not reach statistical significance. While a growing literature shows that compared to girls, young boys are more sensitive to family disruptions across a range of outcomes (e.g., Cooper et al., 2011; DiPrete and Buchmann, 2013), some recent research suggests that paternal incarceration is negatively associated with cognitive skills and likelihood for early grade retention at age nine among both boys and girls (Haskins, 2013; Turney and Haskins, 2014). Thus, while evidence is mounting for the vulnerability of young boys’ to paternal incarceration, future work should continue to explore effects for girls across a range of outcomes and developmental stages. Lastly, comparisons across parent- and child-reports of externalizing and internalizing behaviors illuminate differences in both the perceived magnitude of overall effects of paternal incarceration as well as how impacts by respondent perceptions might vary depending on the gender of the child. Parent reports of these behavioral outcomes produced the largest impacts of paternal incarceration, while child self-reports of their own behaviors showed fewer significant differences and were of smaller magnitude (often nearly half the size). Moreover, if this study relied only on parent-reports, slightly different conclusions by gender would have been made as parent-reports of both externalizing and internalizing problem behaviors for girls with incarcerated fathers reached significance while child self-reports did not. These findings suggest a more nuanced understanding is needed. If we believe children are the most accurate informants of their behavior and skills, and social desirability bias is not a major concern, then it is possible studies that rely solely on parent perceptions of children’s behaviors may be over-estimating impacts of paternal incarceration. Future work comparing agreement of child- and parentreports across a range of outcomes would better inform our understanding of both the lived experiences of children of the incarcerated as well as how adults perceive the wellbeing and skill capacities of this growing group of American children. 4.1. Limitations Future work to assuage additional study limitations is necessary. This study aimed to establish whether paternal incarceration was associated with deleterious consequences for behavioral components of children’s non-cognitive skills in middle childhood; however a test of mechanisms was beyond its scope. Also, only considered were children with first-time paternal incarceration experiences between ages 1 and 9, limiting from the analytic sample a broad swath of children whose exposure was earlier (prior to age 1) but timing was harder to pinpoint. Given this reduction, results are potentially underestimated. Moreover, the decreased sample size for subgroup analyses by gender limit the definitiveness to which conclusions can be drawn as the precision of the point estimates weakens with sample reductions. Additionally, while very little evidence for variation in effects of paternal incarceration by race has been documented, the lack of subgroup analyses by race might be considered a limitation as exposure to paternal incarceration is disproportionately experienced by racial minorities (Western and Wildeman, 2009). Lastly, threats to causal inference remain as matching models cannot fully address omitted variable bias in the same way as an experimental study. While this study’s design featured a number of important elements, including controlling for a large range of variables associated with the main selection concerns, utilizing a reference group similarly at risk of paternal incarceration through propensity score matching, incorporating multiple informants of child outcomes, and featuring appropriate time-ordering of variables, nevertheless, it is possible there exists unmeasured factors associated with increases in children’s antisocial behaviors and paternal incarceration. Future empirical work on the intergenerational effects of paternal incarceration is needed and thus scholars must continue to grapple with such issues. 5. Conclusion The first ten years of life constitute a critical period in childhood for the healthy development of age-appropriate skills. During the developmental periods of early and middle childhood the foundation for one’s cognitive, social and behavioral capacities begins to solidify into relatively consistent patterns of behavior and skill trajectories that persist into adulthood. Disruptions, stress, and instability experienced due to paternal incarceration during this particularly sensitive period not only have short-term implications for children’s socio-emotional development, but also long-term ramifications for future academic attainment and labor market trajectories. As alarming as early skill development is for later success, some reassurance can be found in work that shows socio-emotional and behavioral capacities appear to be quite malleable to social policy (Heckman et al., 2010; Knudsen et al., 2006; Tough, 2012). Thus identifying childhood experiences that might disrupt healthy non-cognitive skill development early can help us craft policies that best meet children’s needs and potentially curtail cyclical transmissions of disadvantage. Findings from this study are in line with recent work suggesting the incarceration of a father presents a significant impediment to a child’s healthy non-cognitive development and thus their future socioeconomic success and mobility prospects. However findings also contribute new knowledge and a nuanced account of the effects of paternal incarceration for child wellbeing and development. Evidence that the deleterious impact of paternal incarceration on behavioral development extends into middle childhood can help inform policy and programs geared toward elementary-aged children, as they comprise the majority of children impacted by parental incarceration. Moreover, null effects on prosocial behavioral skills and differences in patterns between parent and child reports of antisocial behaviors add complexity to our understanding of the effects of paternal incarceration on children. Similar to work emphasizing resilience processes in children of incarcerated parents (e.g. Poehlmann and Mark Eddy, 2013), the finding that paternal incarceration does not appear to be detrimental for

31

A.R. Haskins / Social Science Research 52 (2015) 18–33

a measure of children’s prosocial development suggests a call for more work exploring the potential protective functions of prosocial behaviors for children of the incarcerated. Lastly, differences in findings based on reporting source (parent or child), particularly in magnitude and by gender, highlight the usefulness of multiple respondents and the continued importance of examining impacts across a range of demographic characteristics and age groups. Paternal incarceration is likely a partial but important avenue through which inequality is produced and reproduced among contemporary urban cohorts of young American children. Acknowledgments This research was supported by a Ford Foundation Dissertation Fellowship, the American Sociological Association’s Minority Fellowship Program, a Provost’s Postdoctoral Research Fellowship from Columbia University and grant number R24HD058486 awarded to the Columbia Population Research Center from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD). The author also thanks the NICHD through grants R01HD36916, R01HD39135, and R01HD40421, as well as a consortium of private foundations for their support of the Fragile Families and Child Wellbeing Study. The author would also like to acknowledge Steven Alvarado, Jeff Grigg, Paul Hanselman, Chris Wildeman and Elizabeth Wrigley-Field for advice, comments and encouragement. The opinions expressed are those of the author only. Appendix A Balance statistics for Parent-Reported Socio-Emotional Skills. Mean bias before/after matching

Paternal incarceration

Diff

Matched pairs

Before

After

% Reduction

SE

T-stat

N

t

c

(raw)

(matched)

in bias

CBCL externalizing behavior Overall 0.330⁄⁄ Within boys 0.430⁄⁄ Within girls 0.272⁄

0.078 0.102 0.113

4.239 4.229 2.402

2042 1079 963

519 271 230

1518 799 721

20.5 22.2 19.3

2.8 2.8 3.0

86.3 87.4 84.4

CBCL internalizing behavior Overall 0.174⁄ Within boys 0.177 Within girls 0.229⁄

0.074 0.111 0.114

2.348 1.591 2.00

1988 1054 934

509 258 227

1456 780 696

20.7 22.3 19.6

2.8 2.6 3.4

86.5 88.3 82.7

Notes: Kernel matching model estimates shown. See Table 1 and Methods section for a complete list of variables used in the models predicting the treatment. Analyses are unweighted and done on imputed data. Matched pairs indicate the average number of treated and control observations on common support. Significance levels are the following: ⁄p < .05; ⁄⁄p < .01; ⁄⁄⁄p < .001 (two-sided).

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ssresearch.2015.01.001. References Allen, Sarah, Daly, Kerry, 2007. The Effects of Father Involvement: An Updated Research Summary of the Evidence vol. Ontario. University of Guelph, Center for Families, Work and Wellbeing, Canada. Allison, Paul D., 2001. Missing Data. Sage Publications Inc, Thousand Oaks, CA. Augurzky, Boris, Schmidt, Christoph M., 2001. The Propensity Score: A Means to an End’’. IZA Discussion Paper 271. Institute for the Study of Labor (IZA). Blair, Clancy, 2002. School readiness: integrating cognition and emotion in a neurobiological conceptualization of children’s functioning at school entry. Am. Psychol. 57 (2), 111–127. http://dx.doi.org/10.1037//0003-066x.57.2.111. Bocknek, Erika L., Sanderson, Jessica, Britner, Preston A., 2009. Ambiguous loss and posttraumatic stress in school-age children of prisoners. J. Child Fam. Stud. 18 (3), 323–333. Bowles, S., Gintis, H., 1979. Schooling in Capitalist America. Basic Books, New York. Buchmann, Claudia, DiPrete, Thomas A., McDaniel, Anne, 2008. Gender Inequalities in Education, Annual Review of Sociology, vol. 34, Annual Review of Sociology, pp. 319–37. Caliendo, Marco, Kopeinig, Sabine, 2008. Some practical guidance for the implementation of propensity score matching. J. Econ. Surveys 22 (1), 31–72. Carneiro, Pedro, Heckman, James J., 2003. Human Capital Policy. In: Heckman, James J., Krueger, Alan B., Friedman, Benjamin M. (Eds.), Inequality in America: What Role for Human Capital Policies? MIT Press, Cambridge, MA. Cho, Rosa M., 2010. Maternal incarceration and children’s adolescent outcomes: timing and dosage. Soc. Service Rev. 84 (2), 257–282.

32

A.R. Haskins / Social Science Research 52 (2015) 18–33

Cooper, Carey E., Osborne, Cynthia A., Beck, Audrey N., McLanahan, Sara S., 2011. Partnership instability, school readiness, and gender disparities. Sociol. Educ. 84 (3), 246–259. http://dx.doi.org/10.1177/0038040711402361. Craigie, Terry-Ann L., 2011. The effect of paternal incarceration on early child behavioral problems: a racial comparison. J. Ethnic. Criminal Justice 9 (3), 179– 199. Dallaire, D.H., Zeman, J.L., 2013. Empathy as a protective factor for children with incarcerated parents. Monogr. Soc. Res. Child Dev. 78, 7–25. http:// dx.doi.org/10.1111/mono.12018. Dehejia, Rajeev H., Wahba, Sadek, 1999. Causal effects in nonexperimental studies: reevaluating the evaluation of training programs. J. Am. Statist. Assoc. 94 (448), 1053–1062. http://dx.doi.org/10.2307/2669919. DiPrete, Thomas A., Jennings, Jennifer L., 2012. Social and behavioral skills and the gender gap in early educational achievement. Soc. Sci. Res. 41 (1), 1–15. http://dx.doi.org/10.1016/j.ssresearch.2011.09.001. DiPrete, Thomas A., Buchmann, Claudia, 2013. The Rise of Women: The Growing Gender Gap in Education and What It Means for American Schools. Russell Sage Foundation, New York, NY. Dong, Nianbo, Maynard, Rebecca, 2013. Powerup!: a tool for calculating minimum detectable effect sizes and minimum required sample sizes for experimental and quasi-experimental design studies. J. Res. Educ. Effect. 6 (1), 24–67. http://dx.doi.org/10.1080/19345747.2012.673143. Duckworth, Angela L., Seligman, Martin E.P., 2005. Self-discipline outdoes iq in predicting academic performance of adolescents. Psychol. Sci. 16 (12), 939– 944. http://dx.doi.org/10.1111/j.1467-9280.2005.01641.x. Duncan, Greg J., Dowsett, Chantelle J., Claessens, Amy, Magnuson, Katherine A., Huston, Aletha C., Klebanov, Pamela, Pagani, Linda S., Feinstein, Leon, Engel, Mimi, Brooks-Gunn, Jeanne, 2007. School readiness and later achievement. Dev. Psychol. 43 (6), 1428–1446. Dunifon, R., Duncan, G.J., 1998. Long-run effects of motivation on labor-market success. Soc. Psychol. Quart. 61 (1), 33–48. http://dx.doi.org/10.2307/ 2787056. Farkas, G., 2003. Cognitive skills and noncognitive traits and behaviors in stratification processes. Ann. Rev. Sociol. 29, 541–562. http://dx.doi.org/10.1146/ annurev.soc.29.010202.100023. Farkas, G., Sheehan, D., Grobe, R.P., Shuan, Y., 1990. Cultural resources and school success – gender, ethnicity, and poverty groups within an urban schooldistrict. Am. Sociol. Rev. 55 (1), 127–142. http://dx.doi.org/10.2307/2095708. Foster, H., Hagan, J., 2013. Maternal and paternal imprisonment in the stress process. Soc. Sci. Res. 42 (3), 650–669. http://dx.doi.org/10.1016/ j.ssresearch.2013.01.008. Fritsch, T.A., Burkhead, J.D., 1981. Behavioral reactions of children to parental absence due to imprisonment. Fam. Relat. 30 (1), 83–88. http://dx.doi.org/ 10.2307/584240. Furstenberg, Frank F., Cook, Thomas D., Eccles, Jacquelynne, Elder, Glen H., Sameroff, Arnold (Eds.), 1999. Managing to Make It: Urban Families and Adolescent Success. University of Chicago Press, Chicago, IL. Geller, Amanda, Cooper, Carey E., Garfinkel, Irwin, Schwartz-Soicher, Ofira, Mincy, Ronald B., 2012. Beyond absenteeism: father incarceration and child development. Demography 49 (1), 49–76. http://dx.doi.org/10.1007/s13524-011-0081-9. Haskins, Anna R., 2009. The Effect of Paternal Incarceration on Child School Readiness. Masters Thesis, Sociology, University of Wisconsin-Madison, ProQuest. Haskins, Anna R., 2013. Mass Imprisonment and the Intergenerational Transmission of Disadvantage: Paternal Incarceration and Children’s Cognitive Skill Development. Fragile Families Working Paper Series, No. WP13-15FF. Center for Research on Child Wellbeing: Princeton University. Haskins, Anna R., 2014. Unintended consequences: effects of paternal incarceration on child school readiness and later special education placement. Sociol. Sci. 1, 141–158. Heckman, J.J., Rubinstein, Y., 2001. The importance of noncognitive skills: lessons from the GED testing program. Am. Econ. Rev. 91 (2), 145–149. http:// dx.doi.org/10.1257/aer.91.2.145. Heckman, James J., Stixrud, Jora, Urzua, Sergio, 2006. The effects of cognitive and noncognitive abilities on labor market outcomes and social behavior. J. Labor Econ. 24 (3), 411–482. http://dx.doi.org/10.1086/504455. Hernandez, D.H., Brandon, P.D., 2002. Who are the fathers of the 1990s. Handbook of father involvement. Lawrence Erlbaum, Mahwah, NJ. Heckman, James J., Moon, Seong Hyeok, Pinto, Rodrigo, Savelyev, Peter, Yavitz, Adam, 2010. A New Cost-Benefit and Rate of Return Analysis for the Perry Preschool Program: A Summary. In: N.B.O.E. RESEARCH (Ed.), Working Paper Series, No. 16180, Cambridge, MA. Huston, Aletha C., Ripke, Marika N., 2006. Developmental Contexts in Middle Childhood: Bridges to Adolescence and Adulthood. Cambridge University Press, New York, NY. Jencks, Christopher (Ed.), 1979. Who Gets Ahead?: The Determinants of Economic Success in America. Basic Books, New York. Johnson, Rucker C., 2009. Ever-increasing levels of parental incarceration and the consequences for children. In: Raphael, S., Stoll, M. (Eds.), Do Prisons Make Us Safer? Russell Sage Foundation, New York, pp. 177–206. Johnson, Elizabeth I., Easterling, Beth, 2012. Understanding unique effects of parental incarceration on children: challenges, progress, and recommendations. J. Marriage Family 74 (2), 342–356. http://dx.doi.org/10.1111/j.1741-3737.2012.00957.x. Knudsen, Eric I., Heckman, James J., Cameron, Judy L., Shonkoff, Jack P., 2006. Economic, neurobiological, and behavioral perspectives on building America’s future workforce. Proc. Natl. Acad. Sci. 103, 10155–10162. Kowaleski-Jones, Lori, Duncan, Greg J., 1999. The structure of achievement and behavior across middle childhood. Child Dev. 70 (4), 930–943. http:// dx.doi.org/10.1111/1467-8624.00067. Lamb, Michael E. (Ed.), 2010. The Role of the Father in Child Development. Wiley, New York, NY. Lee, R.D., Fang, X.M., Luo, F.J., 2013. The impact of parental incarceration on the physical and mental health of young adults. Pediatrics 131 (4), E1188– E1195. http://dx.doi.org/10.1542/peds.2012-0627. Marsh, Herbert W., 1990. Sdq Ii Manual: Self Description Questionnaire-Ii. University of Western Sydney, Macarthur, Campbelltown, Australia. Maruschak, Laura M., Glaze, Lauren E., Mumola, Christopher J., 2010. Incarcerated parents and their children. In: Eddy, J.M., Poehlmann, J. (Eds.), Children of Incarcerated Parents: A Handbook for Researchers and Practitioners. Urban Institute Press, Washington, D.C., pp. 33–51. Maumary-Gremaud, Anne, 2000. Things That You Have Done, Grade 4/Year 5 (Technical Report). vol.: FAST Track Project. McLeod, Jane D., Kaiser, Karen, 2004. Childhood emotional and behavioral problems and educational attainment. Am. Sociol. Rev. 69 (5), 636–658. Murray, Joseph, Farrington, David P., 2008. Parental imprisonment: long-lasting effects on boys’ internalizing problems through the life course. Dev. Psychopathol. 20 (01), 273–290. Murray, Joseph, Farrington, David P., Sekol, Ivana, 2012a. Children’s antisocial behavior, mental health, drug use, and educational performance after parental incarceration: a systematic review and meta-analysis. Psychol. Bull. 138, 175–210. http://dx.doi.org/10.1037/a0026407. Murray, Joseph, Loeber, Rolf, Pardini, Dustin, 2012b. Parental involvement in the criminal justice system and the development of youth theft, marijuana use, depression, and poor academic performance. Criminology 50 (1), 255–302. http://dx.doi.org/10.1111/j.1745-9125.2011.00257.x. Myers, B.J., Mackintosh, V.H., Kuznetsova, M.I., Lotze, G.M., Best, A.M., Ravindran, N., 2013. Teasing, bullying and emotions regulation in children of incarcerated mothers. Monogr. Soc. Res. Child Dev. 78 (3), 26–40. National Research Council, 2014. Consequences for Families and Children. The Growth of Incarceration in the United States: Exploring Causes and Consequences. Committee on Causes and Consequences of High Rates of Incarceration, In: Travis, J., Western, B., Redburn, S., (Eds.), Committee on Law and Justice, Division of Behavioral and Social Sciences and Education. Washington, DC, The National Academies Press. Nesmith, Ande, Ruhland, Ebony, 2008. Children of incarcerated parents: challenges and resiliency, in their own words. Child. Youth Serv. Rev. 30 (10), 1119– 1130. http://dx.doi.org/10.1016/j.childyouth.2008.02.006. Poehlmann, Julie, 2005. Representations of attachment relationships in children of incarcerated mothers. Child Dev. 76 (3), 679–696.

A.R. Haskins / Social Science Research 52 (2015) 18–33

33

Poehlmann, Julie, Mark Eddy, J., 2013. Relationship processes and resilience in children with incarcerated parents. Monogr. Soc. Res. Child Dev. 78, 1–6. http://dx.doi.org/10.1111/mono.12017. Reichman, Nancy E., Teitler, Julian O., Garfinkel, Irwin, McLanahan, Sara S., 2001. Fragile families: sample and design. Child. Youth Serv. Rev. 23, 303–326. Roettger, Michael E., Swisher, Raymond R., 2011. Associations of fathers’ history of incarceration with sons’ delinquency and arrest among black, white, and hispanic males in the united states. Criminology 49 (4), 1109–1147. http://dx.doi.org/10.1111/j.1745-9125.2011.00253.x. Rosenbaum, Paul R., Rubin, Donald B., 1983. The central role of the propensity score in observational studies for causal effects. Biometrika 70 (1), 41–55. Royston, P., 2005a. Multiple imputation of missing values: Update. Stata Journal 5, 188–201. Royston, P., 2005b. Multiple imputation of missing values: Update of ice. Stata Journal 5, 527–536. Rubin, Donald B., 1987. Multiple Imputation for Non-Response in Surveys. John Wiley and Sons, New York. Shlafer, R.J., Poehlmann, J., Donelan-McCall, N., 2012. Maternal jail time, conviction, and arrest as predictors of children’s 15-year antisocial outcomes in the context of a nurse home visiting program. J. Clin. Child Adolesc. Psychol. 41 (1), 38–52. http://dx.doi.org/10.1080/15374416.2012.632345. Shonkoff, Jack P., Phillips, Deborah A., 2000. From Neurons to Neighborhoods: The Science of Early Childhood Development. The National Academies Press, Washington, DC. Tough, Paul, 2012. How Children Succeed. Houghton Mifflin Harcourt, New York, NY. Travis, Jeremy, McBride, Elizabeth C., Solomon, Amy L., 2005. Families Left Behind: The Hidden Costs of Incarceration and Reentry, vol. Washington, DC. Turney, Kristin, Haskins, Anna R., 2014. Falling Behind? Children’s early grade retention after paternal incarceration. Sociol. Edu. 87 (4), 241–258. Wakefield, Sara, Wildeman, Christopher, 2011. Mass imprisonment and racial disparities in childhood behavioral problems. Criminol. Public Policy 10 (3), 793–817. http://dx.doi.org/10.1111/j.1745-9133.2011.00740.x. Western, Bruce, Wildeman, Christopher, 2009. The black family and mass incarceration. Ann. Am. Acad. Polit. Soc. Sci. 621, 221–242. http://dx.doi.org/ 10.1177/0002716208324850. Wilbur, M.B., Marani, J.E., Appugliese, D., Woods, R., Siegel, J.A., Cabral, H.J., Frank, D.A., 2007. Socioemotional effects of fathers’ incarceration on low-income, urban, school-aged children. Pediatrics 120 (3), E678–E685. http://dx.doi.org/10.1542/peds.2006-2166. Wildeman, Christopher, 2009. Parental imprisonment, the prison boom, and the concentration of childhood disadvantage. Demography 46 (2), 265–280. http://dx.doi.org/10.1353/dem.0.0052. Wildeman, Christopher, 2010. Paternal incarceration and children’s physically aggressive behaviors evidence from the fragile families and child wellbeing study. Soc. Forces 89 (1), 285–309. Wildeman, Christopher, Turney, Kristin, 2014. Positive, negative, or null? The effects of maternal incarceration on children’s behavioral problems. Demography 51 (3), 1041–1068. http://dx.doi.org/10.1007/s13524-014-0291-z.