Abstract
Prior research on the consequences of imprisonment for the family suggest that incarceration of so-called petty offenders is most harmful for families, yet few studies provide a clear description of who is and is not a petty offender and how best to make such distinctions. We compare various ways of categorizing inmates (using preprison family involvement and characteristics related to criminality and child well-being) to better understand heterogeneity in the consequences of paternal incarceration for children. In our analysis, we find that differentiating between “harmful” and “helpful” fathers is rather difficult, and reform efforts that are overly reliant on criminal offense categories may not be the most gainful policy approach in terms of benefit to children. We also describe a small population of children who appear to benefit from paternal incarceration: children of fathers with severe substance abuse problems. The pattern of results suggests that providing alternative interventions to incarceration, rather than no intervention at all, is critical to improving outcomes for all children of incarcerated parents.
In a relatively short period of time, rigorous research has documented a host of deleterious effects for the children of incarcerated fathers (Foster and Hagan 2007; Geller et al. 2009; Geller, Garfinkel, and Western 2011; Murray and Farrington 2005; Roettger and Boardman 2012; Turney 2014a; Turney and Haskins 2014; Wakefield and Wildeman 2013). The harmful effects of paternal incarceration are comprehensive, influencing everything from mental health and behavioral problems (Geller, Garfinkel, and Western 2011; Wakefield and Wildeman 2011) and delinquency (Murray and Farrington 2005; Roettger and Swisher 2011) to BMI (Roettger and Boardman 2012) to special education placement and early grade retention (Haskins 2014; Turney and Haskins 2014). Several studies also highlight gender (Roettger and Boardman 2012; Wildeman 2010) or race (Craigie 2011; Wildeman 2014) differences in the effects of paternal incarceration on children; others show how the residential status of fathers prior to incarceration (Geller 2013) or family violence (Wildeman 2010) moderates the effect of paternal incarceration.
In prior work with Christopher Wildeman, the first author argued that providing estimates of the average effect of incarceration for children offers a critical baseline for future research on mediators/moderators to be measured (Wakefield and Wildeman 2013). While research on the average effect of paternal incarceration for children is important and useful for some questions, the lack of knowledge on effect heterogeneity means that the body of literature has produced little guidance for policymakers or practitioners to reduce the collateral consequences of incarceration for children. Average treatment effects do not clearly define where practitioners should devote resources to reduce harm. In this article, we focus on a common policy recommendation—reducing incarceration among so-called low-risk or petty offenders—to illustrate the difficulty of translating “average effects” research into useful policy and practical guidance.
Defining Risky Fathers
The evidence suggesting that children of violent fathers are unlikely to be harmed by their parent’s incarceration is relatively strong. What is less clear is which children (with what group of fathers) are most harmed by paternal incarceration. How the criminal justice system defines violence and risk differs from family- or child-centered approaches. Generally speaking, these two frameworks think about risk in fundamentally different ways: the justice system is concerned with risk to the community, whereas child-based systems prioritize risks to the child or family. In the latter model, incarceration itself is a risk to the family because it introduces instability in parental relationships (Apel, this volume; S. H. Andersen, this volume) and creates stress for children (L. Andersen, this volume; Wildeman, Turney, and Yi, this volume). While notions of risk for children driven by parental criminal involvement and parental incarceration are certainly not mutually exclusive, neither is exhaustive on its own; rather, we suggest that criminal justice and family-centered approaches are frequently considered in a vacuum with little dialogue between the two.
The criminal justice system (quite reasonably) defines violent fathers with respect to their criminal conviction history, while those focused on family functioning might be more concerned with fathers who are violent within the home toward their children or partners, whether or not these behaviors result in contact with the criminal justice system. Violence at home and in the community may overlap to some degree (and certainly a public safety perspective would be concerned with both), but each indicator alone is imperfectly suited to capture what is surely wide variation in the quality and character of father-child relationships, violence and instability, and risks to public safety. Moreover, any blunt classification using criminal conviction is unlikely to tell us much about fathers because there is little evidence that criminal offenders specialize in any one activity (Gottfredson and Hirschi 1990). Beyond a small proportion of high-rate violent offenders, the conceptual distinctions between violent and nonviolent, or serious and petty offenders are, at the very least, blurry.
Drug convictions provide another salient example. Reform discussions aimed at lowering the incarceration rate tend to focus on low-level, petty, or, often, drug offenders. This category of offender is thought to present little risk to the general public or to his or her family. However, this broad category of the convicted muddies significant heterogeneity that may not square with empirical findings about crime, risk, and family instability. For example, heavy involvement in the drug trade is not without threat of violence (Blumstein 1995; Reuter 2009), and parental substance abuse presents great difficulties for children and families, whether or not such abuse results in contact with the criminal justice system (Comfort, this volume; Phillips et al. 2006; Tasca, Rodriguez, and Zatz 2011). For those focused on the War on Drugs, punitive sentencing models, or the effects of the prison boom on social inequality, nonviolent drug offenders remain the low-hanging fruit of reform (National Research Council 2014). We do not know, however, whether reducing incarceration among this group would improve family life or childhood well-being.
Finally, it is important to keep in mind that child welfare and criminal justice systems respond very differently to parental violence, drug use, or mental health problems, and these differences likely have large effects on a child’s experience with state intervention (see also Berger et al., this volume). Consider, for example, the response of the child welfare system relative to the criminal justice system to a report of parental violence. While entry into the child welfare system is no panacea for children (Bartholet 1999; Doyle 2007, 2008), children’s best interests are front and center within the social welfare model (at least in theory). In contrast, the criminal justice system is not required to consider the impact of a parent’s incarceration or treatment needs on children; indeed, consideration of these factors would likely violate a number of important sentencing principles, break several laws, and arguably create more problems than it would solve. Because the criminal justice system is (often) legally prohibited from considering the interests of children of incarcerated parents, it is perhaps overly optimistic to expect reforms to criminal justice to result in large well-being gains for children.
There is reason to suspect that even an unintended byproduct of reducing incarceration overall or incarceration of petty offenders specifically might help vulnerable children. The incarceration of “irresponsible but not dangerous” men (e.g., Garfinkel, McLanahan, and Hanson 1998) maps quite nicely onto the nonviolent drug offender who serves in many narratives as the face of incarceration excess. The children of petty offenders, however defined, might be better off if their fathers were not incarcerated. In a recent paper utilizing propensity score comparisons, Kristin Turney (2014b) shows that the harmful effects of paternal incarceration are driven by the children of fathers least likely to be incarcerated, while paternal incarceration is “largely inconsequential” for children of fathers who are at high risk of entering prison, based on a variety of background variables. Work by Turney, Garfinkel and colleagues, and others is useful insofar as it suggests that there exists some, possibly large, population of incarcerated fathers for whom reduced incarceration would benefit their children; more difficult, of course, is figuring out what features or traits are most important in making such distinctions. Whether reduced incarceration helps vulnerable families also depends on what replaces it—at least one study directly focused on children of incarcerated parents suggests that decarceration in the absence of significant investment in treatment programs for parents with substance abuse or mental health problems is unlikely to reduce harm to families (Phillips et al. 2006).
Here, we examine whether it is helpful to account for more background information about incarcerated fathers in analyses of the effects of paternal incarceration on children’s behavioral problems. We thus ask whether factors that are most important for childhood well-being might reasonably be deployed by the criminal justice system when making sentencing, diversion, or treatment decisions; and we assess the potential for criminal justice reform to reduce harm to children of incarcerated parents.
Data and Methods
Data and focal measures
We use data from the Fragile Families and Child Wellbeing Survey (hereafter FFCW; Reichman et al. 2001). The FFCW is a birth cohort study of roughly 5,000 children born in twenty large cities (population over 200,000) in the United States in 1998 and 2000. Mothers and fathers were recruited into the study at the birth of the focal child and followed up at one year, three years, five years, and nine years after birth. The FFCW is an ideal dataset for analyzing the effects of parental incarceration on children—it oversamples nonmarital and disadvantaged families, a group at high risk of experiencing incarceration at some point, and includes gold standard measures of childhood well-being. Since the data collection occurred during a period of high incarceration in the United States, the results are well suited to drawing conclusions about collateral consequences of paternal incarceration during the prison boom era. The utility of the FFCW for studying parental incarceration is borne out in the sheer volume of studies that use it (e.g., Geller et al. 2009, 2012; Geller and Curtis 2011; Haskins 2014; Schwartz-Soicher, Geller, and Garfinkel 2011; Turney 2011, 2014a, 2014b; Turney and Haskins 2014; Turney, Schnittker, and Wildeman 2012; Wakefield and Wildeman 2013; Wildeman 2010; among others).
The current study utilizes data from the first four waves of the survey, both from telephone surveys and in-home assessments, to explore the relationship between paternal incarceration and childhood behavioral problems. Our analytic sample comprises the children of fathers who were incarcerated between the three-year and five-year surveys with available data for pre- and postincarceration measures of behavioral problems, a commonly studied outcome in the literature on children of incarcerated parents. 1 For this analysis, the outcome on which we focus is not a critical concern because we are primarily interested in potential change in effect sizes with the addition of preprison information on incarcerated fathers. Among the available alternatives, though, we chose to focus on aggression for two reasons. First, the harmful effects of paternal incarceration for aggression in their children, especially boys, are among the most studied and robust in the literature (e.g., Geller et al. 2009; Haskins 2014; Wildeman 2010). Second, childhood aggression is among the most predictive of childhood problems for difficulty in adulthood and, importantly, for later criminal involvement (Loeber and Hay 1997; Moffitt 1993). 2
Analytic strategy
Our analysis proceeds as follows. First, using prior studies of the FFCW as a guide (especially Geller et al. 2012; Haskins 2014; Wildeman 2010), we broadly reproduce the effects of paternal incarceration for aggression using simple lagged dependent variable models. These preliminary analyses provide the foundation for our main analysis. We first complete a principal component factor analysis and, using these, group fathers into clusters denoting forms and levels of risk; our aim was to group fathers according to a variety of risk factors that predict both criminal justice involvement as well as problems for their children (more detail on these groups is provided below).
Our analytic technique is essentially a data reduction strategy that allows us to describe how risk factors coalesce within individuals. There are many ways that researchers may do this—in prior work using the FFCW, researchers have tended to adopt a variable-centered approach, including many of the measures we focus on here (for example, impulsivity, mental health problems, or harsh parenting) as regressors. With such a strategy, however, it is difficult to intuitively understand how all of these risks coalesce within individuals. Our models therefore supplement prior research more directly focused on making causal claims by taking a person-centered approach; we use clusters of risk to broadly define fathers with respect to risks prior to entering prison and ask whether they help us to refine our understanding of the relationship between paternal incarceration and childhood behavioral problems. Table 1 provides basic descriptive statistics for the main variables included in our cluster analysis (and Table A1 in the appendix provides descriptive statistics for all additional control variables).
Main Analytic Variables, FFCW
Results
Baseline models
While not the main focus of this article, we note that we were able to reproduce the pattern of paternal incarceration effects on children’s aggression found in prior work (e.g., Geller et al. 2012; Wildeman 2010). 3 Paternal incarceration is associated with large and statistically significant increases in aggression. The effects hold despite the inclusion of numerous controls (measured before the incarceration treatment experience) for socioeconomic status, parent characteristics, and most notably, a prior measure of the dependent variable and prior paternal incarceration. In the full sample, using standardized outcome variables reflecting changes in standard deviations, recent paternal incarceration increases aggression by .17 and, among boys, by .31. In all models, effects are statistically significant at least at the .01 level.
Cluster analysis
Analyses of the FFCW show large and robust effects of paternal incarceration on children’s aggression, especially for boys, but how and for whom are these effects most consequential? In our main analysis, we attempt to add clarity to the discussion by describing the behavior and characteristics of fathers prior to entering prison. We first examine a series of characteristics likely to be consequential for paternal entry into prison and well-being for their children, including mental health problems, substance abuse, domestic violence, self-control, and parenting practices.
Using background information, including parents’ reports prior to paternal incarceration between the three-year and five-year surveys, we completed a principal component factor analysis of fathers in the FFCW that yielded four distinct factors: (1) mental health problems, (2) domestic violence, (3) substance abuse and low self-control, and (4) harsh or erratic parenting. Mental health problems are measured with variables indicating depressive symptoms and whether the father meets the criteria for major depression and/or generalized anxiety disorder. Domestic violence includes whether there is a history of violence between mother and father and mother’s report of violence occurring in front of the focal child. Substance use is indicated by counts of the number of dependency symptoms reported by fathers indicating they had up to four drinks in one day or used an illegal drug in the past six months. These variables are thus an indicator of alcohol and drug abuse, rather than mere use. Self-control is measured by a series of items from Dickman’s (1990) impulsivity scale, measuring the respondent’s ability to carefully deliberate before acting. Finally, harsh and erratic parenting includes (1) continuous measures of physical punishment of the child (hitting, spanking, slapping), (2) shouting, threatening, or cursing at the child, and (3) frequency of father spanking the child. While we would like to isolate behavior that is only related to the father’s parenting, with the exception of paternal spanking, the first and second components of harsh and erratic parenting may include behavior on the part of the father, mother, or other caregiver.
As noted previously, most (if not all) of the variables used in our factor analysis have been included as controls in prior research, but few studies have focused explicitly on how the collection of preprison father characteristics coalesce to produce variation in the effects of paternal incarceration on children. We first identified distinct facets of father characteristics through our factor analysis, then grouped fathers with respect to variation across the resulting four factors. The factor analysis yielded few surprises—mental health problems emerged as distinct from measures of violence, substance abuse emerged as distinct from measures of parenting style—with one exception: Low self-control, a strong positive predictor of criminal involvement as well as substance abuse and other reckless behaviors (Gottfredson and Hirschi 1990; Pratt and Cullen 2000), did not emerge as a separate risk factor nor did it “load” particularly strongly with any other risk factor. It was most strongly related to substance abuse so, given its substantive importance as a predictor of crime, we elected to include it there rather than remove it from consideration entirely.
Using the risk factors described above, we then clustered fathers with respect to their levels on each risk factor. Our cluster analysis yielded five clearly defined groups of fathers that vary with respect to the presence and level of violence, substance abuse/low self-control, mental health problems, and harsh/erratic parenting; these groups are displayed in Figure 1, with each risk factor displayed in the legend. All factors are standardized, and the figure displays the average distance of each group from the mean of each factor. In other words, a +1 on mental health problems indicates that the group includes fathers whose average value on the mental health problems factor is one standard deviation higher than the sample mean. 4 The five groups include the following: (1) a group that is lower than average on all factors (“Low Risk,” n = 1,010), (2) a group of violent fathers who are at the mean on everything else (“Domestic Violence,” n = 164 ), (3) a group of moderately harsh and/or erratic parents who are at the mean on everything else (“Harsh Parenting,” n = 990), (4) a (small) group of fathers with significant substance abuse problems, low self-control, and comorbid mental health problems (“Substance Abuse,” n = 88), and (5) a group of fathers with significant mental health problems and at the mean on everything else (“Mental Health Problems,” n = 185). Most fathers fall into the broadest categories, that of low-risk fathers and harsh/erratic parents. Fewer fathers cluster in the categories that denote rare and severe risks for their children—less than 10 percent of men fall into the domestic violence and mental health problems category and less than 5 percent fall into the substance abuse/low self-control category.
Cluster Membership of Fathers by Risks, FFCW
Before proceeding to a multivariate analysis including cluster membership, it is instructive to examine how well (or not) the clusters foreshadow future incarceration or conviction crime type. Figure 2 displays incarceration between the three-year and five-year follow-ups of the FFCW by cluster membership at the three-year survey. Comparing across cluster types and the likelihood of incarceration, smaller percentages of men from the “Low Risk” category (13 percent) and “Harsh Parenting” category (16 percent) at the three-year survey are incarcerated by the five-year survey relative to those in the “Domestic Violence” (44 percent), “Substance Abuse” (39 percent), and “Mental Health Problems” (24 percent) clusters. Descriptively at least, men with more severe problems, particularly violence and substance abuse, are more likely to be incarcerated. Regression models 5 including father demographics and neighborhood controls also suggest that cluster membership in the domestic violence and substance abuse categories is significantly correlated with later paternal incarceration (OR = 1.91 and 2.72, respectively, relative to membership in the “Low Risk” cluster).
Cluster Membership (%) by Incarceration Status, FFCW
Still, of the 430 men who were incarcerated between the three- and five-year surveys, roughly 29 percent fell into the Low Risk cluster and another 36 percent fell into the Harsh Parenting cluster relative to 17 percent, 8 percent, and 10 percent in the Domestic Violence, Substance Abuse, and Mental Health Problems groups, respectively. This means that the vast majority of fathers in the FFCW who were incarcerated during this time evidenced no glaring red flag just before their incarceration, at least with respect to the risk factors that we consider in this analysis. Cluster membership also does little to distinguish fathers in terms of criminal conviction type, though we caution that this information is available for only a small group of incarcerated fathers and is likely subject to measurement error (see Geller, Jaeger, and Pace, this volume). Figure 3 displays the distribution of incarcerated fathers for whom crime conviction data are available across clusters with respect to the most serious criminal offense that led to incarceration. 6 Most fathers are incarcerated for neither drug nor violent crimes, and cluster membership does not map on particularly well to conviction offense. For example, even though fathers with significant substance abuse problems are more likely to be incarcerated relative to the other clusters, they are not more likely to be incarcerated for drug crimes. Put simply, there is little overlap between being a “risky” father (at least with respect to the characteristics that we examine here) and the seriousness or type of criminal conviction that leads to incarceration.
Cluster Membership (%) by Most Serious Conviction Offense, FFCW
Multivariate analysis using cluster membership
With these descriptive differences in mind, we reestimated our baseline paternal incarceration models to include interaction terms for paternal incarceration and cluster group. The results for aggression are displayed in Table 3 for the full sample and boys only. The table presents (1) the main effect of cluster membership, (2) the main effect of paternal incarceration, (3) the interaction between paternal incarceration and cluster membership, (4) the effect of paternal incarceration within each cluster, and (5) Cohen’s D calculations of substantive significance of the interaction term. Each model includes a prior measure of the dependent variable so the coefficients can be interpreted as the change in behavioral problems associated with each covariate. 7
Baseline Models, FFCW
NOTE: Estimates include the following list of control variables (see Table A1): father’s age (centered); mother’s age (centered); mother and father’s education: Bachelor’s degree, some college, and high school diploma (reference group = less than high school); days father spent with child in past month; mother has new partner; mother and father’s relationship quality; any domestic violence; mother smoked during pregnancy; child’s race; child born in poverty; use of corporal punishment; harsh/erratic parenting within household; number of siblings; mother and father’s self-control; maternal mastery; neighborhood disorder; collective efficacy. All controls measured at the three-year wave or earlier.
p < .01. ***p < .001.
Cluster Analysis, FFCW
NOTE: Included control variables: father’s prior incarceration (up to three-year survey); prior aggression (standardized); mother’s and father’s age (centered); mother’s and father’s education: bachelor’s degree, some college, high school diploma (reference = less than high school diploma); number of days father spent with child last month; mother has new partner; mother smoked during pregnancy; child has low birth weight; child’s race (reference = black); in poverty at child’s birth; number of siblings; mother’s self-control; maternal mastery; neighborhood disorder; collective efficacy.
p < .10. *p < .05. **p < .01.
We begin by noting several important caveats. First, the small sample size of incarcerated fathers and missing data severely limit our ability to detect statistically significant effects. To give a sense of the substantive significance of the effects, we calculate Cohen’s D, an alternative measure often used in contexts of small sample sizes. Second, many of the cluster groups are likely endogenous to incarceration, so we avoid making strictly causal arguments here. Finally, we remind readers of the pattern of change in aggression from Table 1; the pattern in the full sample and among boys is in the direction of less aggression over time. This makes good sense given the age range of children in the sample; the children range in age from three to five, precisely when we might expect them to lessen such behavior. Because the models include a prior measure of the dependent variable, the results should be interpreted as the change in aggression from age three to age five.
Beginning with panel 1, the results suggest that, relative to the low-risk group, the children of nonincarcerated fathers in all other clusters are slower to “age out” of aggression. Only one effect is statistically significant, however; the children of nonincarcerated fathers with severe substance abuse problems have significantly more aggression problems relative to the children of low-risk fathers. Moving to panel 2, the next section of results provides the effect of paternal incarceration for children of low-risk fathers. This effect is positive, reflecting more aggression among children of incarcerated fathers, but the effect is statistically significant only for boys.
Turning to the interaction between paternal incarceration and cluster membership, in panel 3, the results are both expected and surprising. Using statistical significance as a guide, the most notable effect is with respect to children of fathers with severe substance abuse problems (and recall that this group of fathers also exhibited relatively high comorbid mental health problems and relatively low self-control). While both severe substance abuse and paternal incarceration are generally associated with more aggression problems, the interaction between the two is in the opposite direction—suggesting that behavioral problems are lower among children with incarcerated fathers who have severe substance abuse problems relative to children of incarcerated fathers in the low-risk cluster. Among children who have fathers with severe substance abuse problems, aggression is much higher for those whose fathers are not incarcerated.
The next section of the panel (4) provides the effect of paternal incarceration on aggression within cluster groups. It simply adds the main effect of incarceration with the interaction effects (e.g., for substance use, total sample: .145 + -.679 = -.534). For all clusters except the substance abuse group and consistent with prior research, the effect of paternal incarceration on aggression is higher than the sample mean. The exception again are children with fathers who have severe substance abuse problems—for these children, aggression is the lowest relative to all other groups. Among this small group of children, paternal incarceration appears to be beneficial.
The results of the cluster analysis are limited by small sample sizes and issues of endogeneity as well as limited to fathers who experienced incarceration between only two waves of FFCW data collection. Viewed in that light, it is perhaps unsurprising that few of our comparisons rise to standard levels of statistical significance. We therefore conclude our analysis with a comparison of the substantive significance of the results. The bottom panel of Table 3 displays measures of substantive significance, Cohen’s D, for the effect of paternal incarceration by cluster membership interactions. The size of the estimate gives a rough sense of the importance of the effect as follows: .2 or less = not noteworthy, .5 = medium effect, .8 or above = large effect. The results of Cohen’s D underscore those in the rest of the table; the effect of paternal incarceration for the children of fathers with severe substance abuse problems is a meaningful and important departure from the others in both size and direction.
Conclusion
Quantitative research on children of incarcerated fathers has tended to mask significant heterogeneity in children’s experiences (see also L. Andersen, this volume). Using prior work as a guide, we offered evidence that background characteristics of fathers do not reliably predict the likelihood of incarceration, except in the case of fathers with severe substance abuse problems and severe mental health problems. That a third of incarcerated men clustered in the “low risk” category is problematic and, consistent with narratives of the rise of the prison boom, suggests that the harmful consequences of incarceration are partly driven by the incarceration of men who present little danger to their families (and, of course, data limitations that prevent us from properly categorizing men with regard to risk). And yet that is only part of the story. Our analysis shows that low risk is not synonymous with a nonviolent drug offender—of the 76 low-risk men in this study for whom conviction data are available, one-third were incarcerated for violent crimes. By the same token, self-reported addiction does not reliably predict incarceration for drug crimes; very few of these men were incarcerated for drugs, and most landed in the “other” crime category.
Turning from low- to high-risk fathers, we suggest that the emphasis on the excesses and punitive nature of sentencing policy that drove the prison boom obscures a troubling issue: there are some parents who pose significant problems for their children, either because they suffer from serious mental health problems or because they struggle with addiction. Our results show a small group of children for whom paternal incarceration appears to provide stability, rather than reduce it. These results are in line with other descriptions of incarceration as a period of potential respite for inmates and their families (Comfort 2008 but see Comfort, this volume, for important caveats). While there is ample evidence that prison is an inefficient and ineffective intervention for such parents (National Research Council 2014), they do require some intervention if the goal is to improve childhood well-being while reducing incarceration. The former assertion is more widely recognized today, but there remains little discussion of what precisely should take the place of incarceration, and this is more broadly complicated by large racial disparities in incarceration rates (Haskins and Lee, this volume).
Our results thus present a puzzle for criminal justice reform. Across-the-board reductions in incarceration may improve well-being for some children, but absent alternative interventions, some children may also be harmed. At the very least, our results suggest that criminal justice reform targeted at broad classes of criminal convictions (whether they are drug convictions, petty offenders, nonviolent offenders, or something else) is unlikely to substantially improve well-being for children. The class of inmates convicted of drug offenses, for example, is simply too heterogeneous. The same is true of those convicted of violent offenses. Restricting reform to nonviolent offenders to reduce incarceration rates makes little sense given that 54 percent of state prison inmates are incarcerated for violent crimes (Carson 2014); our results also suggest that the children of violent offenders may not benefit from their incarceration. The result likely reflects heterogeneity among violent offenders as well as deficits in the measures available in survey data. The FFCW (or any other U.S.-based survey) simply does not allow us to make fine-grained distinctions between persistently violent men and those who may have committed one act of violence but who are otherwise committed fathers.
We end our discussion by noting two complications for future research. Largely because of data limitations (see Wakefield, Lee, and Wildeman, this volume; Geller, Jaeger, and Pace, this volume, for more detail), using current research knowledge to produce policy guidance still requires some conceptual sleight of hand. Because of such limitations, we often use evidence on outcomes for children to make recommendations about what is good for their parents or conflate a parent’s desire to remain unincarcerated with the benefit a child might incur from their parent remaining in the home. Yet the interests of (overincarcerated) parents may not necessarily be aligned with the best interests of their children in the absence of alternative interventions. An overall reduction in incarceration coupled with targeted diversion programs that offer intensive treatment to people with severe addiction and/or mental health problems is a fruitful avenue for criminal justice reform. We should move away from reform efforts focused on specific classes of inmates (drug offenders, petty criminals, and the like) and toward policies that tackle the constellation of problems, whether addiction, mental health, or violence, that inmates (and their families) face.
Footnotes
Appendix
Additional Control Variables, FFCW
| Total Sample (N = 1,743) |
Boys (N = 904) |
|||
|---|---|---|---|---|
| Mean | SD | Mean | SD | |
| Father age (centered) | −.225 | 7.159 | −.319 | 7.129 |
| Mother age (centered) | .349 | 6.062 | .493 | 6.142 |
| Child male | .512 | .500 | 1 | 0 |
| Child race | ||||
| Black | .460 | .499 | .463 | .499 |
| White | .252 | .435 | .254 | .436 |
| Hispanic | .255 | .436 | .243 | .429 |
| Other | .033 | .178 | .039 | .193 |
| Maternal education | ||||
| Less than HS | .301 | .459 | .304 | .460 |
| HS diploma | .303 | .460 | .306 | .461 |
| Some college | .263 | .441 | .256 | .436 |
| BA | .133 | .339 | .134 | .341 |
| Paternal education | ||||
| Less than HS | .318 | .466 | .331 | .471 |
| HS diploma | .334 | .472 | .344 | .475 |
| Some college | .226 | .418 | .205 | .404 |
| BA | .121 | .326 | .121 | .326 |
| Days father spent w/child, last 30 days | 21.885 | 12.238 | 21.646 | 12.333 |
| Mother has new partner | .084 | .278 | .087 | .283 |
| Relationship quality (1–5) | 3.336 | 1.353 | 3.360 | 1.364 |
| Mother smoked while pregnant | .188 | .391 | .181 | .386 |
| Child low birth weight | .098 | .298 | .091 | .287 |
| Number of siblings | 1.114 | 1.280 | 1.117 | 1.373 |
| In poverty at child’s birth | .317 | .465 | .330 | .470 |
| Maternal self-control (1–4) | 2.994 | .609 | 2.966 | .626 |
| Maternal mastery (1–4) | 2.749 | .658 | 2.725 | .663 |
| Neighborhood disorder (1–4) | 1.775 | .866 | 1.776 | .864 |
| Collective efficacy (1–5) | 3.075 | .603 | 3.082 | .594 |
NOTE:
We thank Christopher Wildeman and Kristin Turney for graciously answering our questions regarding coding and analytic decisions for the Fragile Families and Childhood Wellbeing Study; and Lars Andersen, Robert Apel, Megan Comfort, Amanda Geller, Hedwig Lee, Steve Raphael, Bill Sabol, and the anonymous reviewer for helpful comments.
Notes
Sara Wakefield is an assistant professor in the School of Criminal Justice at Rutgers University. Her research interests focus on the consequences of mass imprisonment for the family, with an emphasis on childhood well-being and racial inequality. She is the coauthor of Children of the Prison Boom: Mass Incarceration and the Future of American Inequality (with Christopher Wildeman; Oxford University Press 2013). Some of her related work examines the social networks and conditions of confinement of inmates and social/family ties during reentry.
Kathleen Powell is a graduate student in the School of Criminal Justice at Rutgers University. She is currently completing research on the collateral consequences of mass incarceration for children. Her primary interests lie in the transition to adulthood for youth involved in the juvenile justice system.