Abstract

Other researchers have posited that important events in men’s lives—such as employment, marriage, and parenthood—strengthen their social ties and lead them to refrain from crime. A challenge in empirically testing this hypothesis has been the issue of self-selection into life transitions. This study contributes to this literature by estimating the effects of an exogenous life shock on crime. We use data from the Fragile Families and Child Wellbeing Study, augmented with information from hospital medical records, to estimate the effects of the birth of a child with a severe health problem on the likelihood that the infant’s father engages in illegal activities. We conduct a number of auxiliary analyses to examine exogeneity assumptions. We find that having an infant born with a severe health condition increases the likelihood that the father is convicted of a crime in the three-year period following the birth of the child, and at least part of the effect appears to operate through work and changes in parental relationships. These results provide evidence that life events can cause crime and, as such, support the “turning point” hypothesis.

Introduction

Criminal activity is costly to both individuals and society. In the United States, crime resulted in a total loss of almost $16 billion to victims in 2004; and expenditures for police, the criminal judicial system, and corrections exceeded $185 billion in 2003 (The Cost of Crime 2006). The indirect costs of crime also appear to be substantial. For example, children with imprisoned fathers have high rates of poverty, are more likely to grow up without a father, and are often stigmatized, all of which can limit their future labor market opportunities in the legal sector and increase the likelihood that they, like their fathers before them, will engage in criminal activity (Wildeman 2009). For all these reasons, it is important to understand the processes that shape individuals’ criminal behavior.

According to Freeman (1996), crime is an activity of individuals with low legitimate earnings prospects, and “. . . involvement with the criminal justice system has . . . become part of normal economic life for many young men” (p. 25). For example, more than one-half of 25- to 34-year-old black men who are high school dropouts are incarcerated, on probation, or on parole (Freeman 1996). Yet, many disadvantaged men do not commit crimes, and between the ages of 25 and 35, many offenders either decrease their level of criminal activity or desist altogether (Kerner 2005).

Sampson, Laub, and others have posited that important events in men’s lives—such as employment, marriage, and parenthood—strengthen their social ties and lead them to refrain from crime (e.g., Sampson and Laub 1990). Such “turning points” may explain the observed declines in criminal behavior during the life course and could even explain the take-up of crime. Studies have found empirical evidence consistent with the “turning point” hypothesis (e.g., Sampson et al. 2006; Laub et al. 1998; Horney et al. 1995). However, an acknowledged challenge in the literature has been the issue of causality. Life events, such as marriage and employment, are not random; rather, individuals choose these activities just as they choose to commit crime. Thus, it is difficult to say with any degree of certainty that such events cause changes in criminal behavior.

In this study, we exploit a life shock—one that is considered random in the population and has been shown to affect ties to both romantic partners and the labor market—to test the turning point hypothesis. Specifically, using data from the Fragile Families and Child Wellbeing Study that have been augmented with data from medical records, we estimate the effects of the birth of a child with a severe health problem on the likelihood that the infant’s father becomes or remains involved in illegal activities. The shocks that we study are infant health conditions that are considered by the medical community to be random events with poor long-term prognoses. We conduct a number of auxiliary analyses to support our assumption that the conditions in question are, in fact, exogenous. The study sample consists of relatively young and poor urban men, the very group on the margin for committing crime. We explore both desistance from and entry into crime and estimate the potential mediating roles of relationship and labor market ties. We consider cohabiting as well as marital relationships because the former are quite relevant for the population of interest but have rarely been considered in studies of the effects of turning points on criminal behavior.

Background

Pioneering work by Sampson and Laub (1990); Laub and Sampson (1993); Laub et al. (1998), and Sampson et al. (2006) provides a strong theoretical underpinning for examining the impact of adult experiences on crime. The authors posited that institutions of informal social control, such as family and work, shape an individual’s social behavior, and that life events that strengthen ties to those institutions can reduce criminal activity or divert individuals from a criminal career. Conversely, a weakening of ties can lead to increases in criminal activity. The authors referred to life events that shift the trajectory of a person’s pro- or anti-social behavior as “turning points.” Although the turning point hypothesis can potentially apply to many different types of life events, much of the relevant empirical work to date has focused on marriage.1

Sampson et al. (2006) recently highlighted the issue of causality as it relates to estimating the effects of marriage on criminal behavior: “Marriage is not a random event. . . .To the extent that marriage is influenced by individual self-selection, the marriage-crime relationship is potentially spurious” (p. 470). To address this issue, the authors investigated whether marriage reduced crime in a sample of 500 male delinquents born between 1923 and 1930 and followed them to age 70 (Sampson et al. 2006). Using an inverse probability of treatment weighting (IPTW) methodology and conducting numerous specification and robustness tests, they found high magnitude reductions (about 40% to 42%) in crime subsequent to marriage. These results suggest that the marriage-crime relationship found in previous studies (e.g., Horney et al. 1995; Laub et al. 1998; Laub and Sampson 1993; Sampson and Laub 1990) is not spurious. However, the authors concluded with caveats. First, although the IPTW approach has advantages, it would be helpful to apply other approaches that do not rely on the same types of assumptions. Second, the sample they studied is not a contemporary one, and the findings may therefore not be generalizable to the current cohort of young men who are likely to commit crime.

King et al. (2007) used a propensity score matching methodology to estimate the effect of being married on crime in a national sample (from the National Youth Survey) of men aged 21 to 27 in 1987. They found that married men were less likely than unmarried men to engage in crime, but acknowledged that their technique was not an ideal strategy for addressing the causality issue because it cannot take into account unmeasured differences between individuals that may impact both crime and marriage. In other words, their method is an imperfect substitute for a study that has random assignment of the “treatment.”

Two recent analyses used “natural experiment” approaches to investigate causality in the marriage-crime relationship. Edlund et al. (2007) investigated the effects of China’s one-child policy on crime using aggregate data for provinces in China between 1988 and 2004. From difference-in-difference models, the authors found that (1) the one-child policy increased the sex ratio in China, so that by 2004, there were about 1.1 males per female in the 16–25 age cohort compared with a ratio of less than 1.06 in 1986; (2) a 0.01 increase in the sex ratio increased both violent and property crime rates by about 3%; and (3) the increases in crime were due not only to the relative increase in the male population but also to declining marriage rates among men. Caceres-Delpiano and Giolito (2008) examined the impact of unilateral divorce laws in the United States on state-level aggregate arrest rates between 1965 and 1998. Using difference-in-difference models, the authors found evidence that unilateral divorce laws increased arrests for both violent and property crimes among men aged 20 to 24 years. Thus, the Edlund et al. and Caceres-Delpiano and Giolito studies found that policies that decrease marriage among men also lead to increases in crime. These two studies show that natural experiments, such as policy changes, can be used to study the question of whether marriage affects crime. When examining policies over a long period of time, however, other trends could potentially explain some of the observed relationship. For example, Edlund et al. acknowledged that China experienced dramatic societal changes during their sample period, and Caceres-Delpiano and Giolito studied a three-decade period in the United States that was characterized by multifaceted societal change. Additionally, evidence from natural experiments tends to be limited in terms of generalizability. For these reasons, additional tests of the turning point hypothesis are warranted.

As discussed by Sampson et al. (2006), a drawback of their own study was that the sample of men were born well before World War II. Recent cohorts of men behave quite differently than this earlier cohort in regard to two important life turning points: marriage and parenthood. In 1970, less than 20% of men aged 25 to 29 had never been married (Saluter 1994: Table B); that figure increased to 60% in 2008 (U.S. Census Bureau 2009: Table A1). According to Edin and Reed (2005), poor men and women are about one-half as likely to be married as those with incomes three or more times the poverty level. As marriage has declined, nonmarital cohabitation and childbearing have increased. The ratio of unmarried-couple households to 100 married couples increased from 1 in 1970 (Saluter 1994: Table B) to 11 in 2008 (U.S. Census Bureau 2009: Table A1). In 1970, about 10% of all births in the United States were to unmarried women; by 2006, that share increased to 38.5% (National Center for Health Statistics 2011). Recent evidence indicates that a sizable fraction of parents with nonmarital births are in cohabiting relationships; for example, about one-half of unmarried women giving birth in urban hospitals in the United States were cohabiting with the father of the baby at the time of the birth (McLanahan et al. 2003). These recent trends indicate that the marital dichotomy has become less appropriate than it once was for characterizing partnering among poor young men—those who are at high risk for criminal careers.

Capaldi et al. (2008) investigated the effect of romantic partner relationship stability on criminal behavior using a sample of at-risk men from the Oregon Youth Study who were in a partner relationship. They found that relationship stability appeared to decrease criminal activity among men with prior arrests but not among those without prior arrests, and that a measure of the man’s attachment to his partner was not associated with changes in his criminal activity. Although this study broke new ground by considering different dimensions of partnering, it has two key drawbacks. It did not address the causality issue in as rigorous a fashion as the studies described earlier, and the sample size is small.

Another relevant study estimated the effects of entering marital and cohabiting unions on young adults’ substance use. Specifically, Duncan et al. (2006) used the 1979 National Longitudinal Survey of Youth to examine binge drinking, marijuana use, and cigarette smoking before and after relationship formations. They found evidence of significant declines in binge drinking and marijuana use following the formation of marital, but not cohabiting, relationships. That is, they found that compared with the formation of marital unions, the formation of cohabiting unions was less favorably linked to men’s behavior.

In this study, we examine the impact of a life event—namely, the birth of a seriously unhealthy child—on criminal activity in a recent population-based cohort of men who are at high risk for criminal behavior by virtue of their age and socioeconomic status. We exploit our ability to characterize poor child health as a shock in order to address the causality issue. We know from previous research that having an infant born in poorer health than expected reduces both the likelihood that the father lives with the child’s mother (Reichman et al. 2004) and the father’s labor supply (Noonan et al. 2005) one year after the child’s birth. In other words, the birth of an unhealthy infant has been shown to have negative effects on relationship and labor market ties. As such, investigating the effects of this life event on fathers’ criminal activity represents a credible test of the turning point hypothesis.

Data

We use data from a national longitudinal birth cohort survey that have been linked to medical records of the mother respondents and their newborns. The Fragile Families and Child Wellbeing Study (or simply, Fragile Families) follows a cohort of parents and their newborn children in 20 large U.S. cities (in 15 states).2 Births were randomly sampled from 75 hospitals in the 20 cities between 1998 and 2000. By design, approximately three-quarters of the mothers in the sample were unmarried. Face-to-face interviews were conducted with 4,898 mothers while they were still in the hospital after giving birth. The infants’ fathers were also interviewed shortly thereafter, in the hospital or at another location. Baseline response rates were 86% among eligible mothers and 78% among eligible fathers. (Fathers were eligible if the infant’s mother completed an interview.) Additional data have been collected from the hospital medical records (from the birth) for a subsample of 3,684 births.3 Measures of census tract-level poverty were constructed by using the parents’ baseline addresses, whether they lived together or apart.

Follow-up interviews were conducted when the child was 1 and then 3 years old. Eighty-nine percent of the mothers who completed baseline interviews were reinterviewed when their children were between 12 and 18 months old; 86% of mothers who completed baseline interviews were reinterviewed when their children were about 3 years old (between 30 and 50 months). Of the 3,830 fathers who completed baseline interviews, 82% completed one-year follow-up interviews. Of the 3,830 fathers, 77% completed three-year follow-up interviews.

The enhanced Fragile Families data are well suited for analyzing the effects of a life shock on the father’s criminal activity because they were collected as part of a longitudinal birth cohort study and include (1) measures of an adverse life shock—namely, the birth of a child with a serious health condition; (2) information about the father’s criminal activity both prior to and following the shock; and (3) rich data on the father’s and family’s pre-shock characteristics as well as changes in their circumstances following the shock.

Model

Based on the economic framework pioneered by Becker (1968) and Ehrlich (1973), crime is a function of the expected costs and benefits of engaging in that activity and can be modeled as follows:
formula
(1)

Individuals with high levels of human capital (and therefore favorable prospects in the legal labor market) are less likely than those with low levels of human capital to engage in crime. The residential environment is also important; strict criminal justice sanctions deter crime, and weak local labor markets increase crime (e.g., Gould et al. 2002; Grogger 1998; Mocan and Rees 2005). Based on the turning point literature described earlier, Sickles and Williams (2008) incorporated social capital—ties to community through work and marriage—into the economic model. According to Sampson et al. (2006), there are three ways that social ties could deter (men’s) criminal activities. First, they may increase both the opportunity cost of serving time and the disutility from engaging in illegal acts (by increasing the quality of time spent outside of prison or jail). Second, they may increase the opportunity cost of time spent in illegal activities (through increased time obligations to family or work). Third, spouses and partners can play a direct role by monitoring the individual’s behavior and decreasing access to illegal activities.

Our key variable of interest is an adverse life shock—the birth of a child in poorer health than was expected—which, as discussed earlier, has been shown to reduce the father’s ties to both the child’s mother and the legal labor market. As such, we would expect the shock to increase the likelihood of crime, particularly among men who are near the margin for engaging in criminal behavior—that is, those with low levels of human capital, weak social bonds, or a previous criminal record (which can affect both human and social capital).

Sample, Measures, and Descriptive Analysis

Fathers were included in the sample no matter which of the survey waves they completed, as long as medical record data (which were needed for our measures of poor infant health, as described later) were available for that birth and information was available for all analysis variables. Of the 3,684 cases with medical record data, we dropped 63 because the birth was a multiple, 30 because the father was incarcerated at all three survey waves, and 151 because the father had not seen the child at least once during the three years, leaving a potential sample of 3,440 fathers.

Mothers provided much of the relevant baseline information (e.g., demographics) for fathers who did not complete baseline interviews. For each of the three survey waves, we include dummy variables for the father not completing that interview. In general, observations without complete data on all right-hand-side variables were dropped. However, we include a dummy variable for fathers who are missing information on their baseline census tract, regardless of whether they completed a baseline interview.

In the sections that follow, we describe the measures we use in our analyses, present summary statistics, and point out many salient characteristics of the sample.

Criminal Activity

It is notoriously difficult to accurately measure an individual’s criminal activity. Administrative crime records are rarely linked to other individual-level data, so much of what is known about the criminal behavior of individuals comes from surveys. Ideally, individuals would truthfully and accurately recall every crime committed. However, because they may be reluctant to admit to having engaged in serious illegal activities, especially if they have not yet been caught or convicted, individuals’ self-reports of criminal activity may be biased. Many surveys ask about number of contacts with the criminal justice system (incarcerations or convictions), which may be less biased than self-reports of actual criminal activity. Even these measures likely underreport criminal activity, however, because most crimes do not result in arrest, not all arrests result in convictions, and many convictions do not result in incarceration.4

Our two main outcome measures are whether the father was ever convicted of a crime from the time that the focal child was born until three years later and whether he was ever incarcerated during that same three-year period.5 Although we have the timing of both convictions and incarcerations in our data (discussed later), neither measure allows us to precisely pinpoint the timing of the actual crime. (We explore this issue later, to the extent possible.) The conviction measure has the advantage that it is closer to the crime in the crime-arrest-conviction-incarceration process. Advantages of the incarceration measure are that (1) previous research has validated self-reports of incarceration (Lochner and Moretti 2004); (2) our data allow us to validate the father’s self-reported incarceration with the mother’s report of his incarceration history (as described later); and (3) incarceration is a significant outcome in its own right (other than as a proxy for crime) because it involves physical removal of the father from his family, which could affect maternal and child well-being even if the parents are no longer romantically involved (e.g., an incarcerated father is unlikely to pay child support or spend quality time with the child). Thus, both measures have advantages and disadvantages as proxies for crime, and using the two different measures can serve as a validation of our findings. In particular, because not all convictions result in incarceration and because incarceration is more temporally removed from the actual crime, we would expect stronger effects for conviction than for incarceration.

For each outcome, we restrict the sample to cases with nonmissing data on that measure of criminal activity (both pre- and post-birth) and covariates. Of the 3,440 cases potentially in the sample, 2,603 were used for the analyses of conviction (738 were dropped for missing information on convictions after the child’s birth; 15, for missing information on convictions before the birth; and 84, for missing information on one or more covariates), and 2,677 were used for the analyses of incarceration (568 were dropped for missing information on incarceration after the child’s birth; 96, for missing information on incarceration before the birth; and 99, for missing information on one or more covariates). Descriptive statistics are presented in Table 1 for the two analysis samples.6

In the one-year and three-year follow-up surveys, fathers were asked whether they had ever been convicted of a serious crime and, if so, when. Overall, 22% of the fathers reported that they had been convicted of a crime either before or after the birth of the focal child (not shown in tables). Among those who reported the type of crime, the three most prevalent types were drug crimes (37% of the fathers who were ever convicted), property crimes (27%), and violent crimes (23%).7 From the father’s responses, we constructed a variable for whether the father was ever convicted of a crime after the birth of the child and another for whether the father was ever convicted of a crime before the child was born. We use the latter as a control in multivariate analyses of conviction.8 As shown in Table 1, 9% of the fathers were convicted of at least one crime in the three years following the birth of the child, whereas 19% had been convicted of at least one crime before the child was born.

In the one-year and three-year follow-up surveys, both fathers and mothers were asked whether and when the father had ever been incarcerated. We used information provided by the parents to construct measures of fathers’ post- and pre-birth incarceration; we use those measures in multivariate analyses of incarceration.9 As shown in Table 1, 12% of the fathers were incarcerated at some point after the birth of the child, and 14% had been incarcerated some time before the child was born.10

Infant Health Shocks

We exploit our ability to characterize infant health shocks with our data. Unlike most studies that ascertain infant health through survey questions to mothers, we constructed measures of poor infant health based on data from hospital medical records (from the birth) as well as maternal reports from the one-year follow-up survey of specific child disabilities, such as cerebral palsy, that were likely present at birth. The ideal measure of poor infant health would have two characteristics: it would be unrelated to maternal prenatal behavior, and it would be associated with long-term morbidity. Thus, it would be a true exogenous shock, as are many congenital conditions. Here, and in the Appendix, we describe how we constructed the different measures of poor infant health. These range from a very strict measure that approximates the ideal measure but characterizes few infants in our sample as unhealthy, to broader measures that yield a greater number of cases with poor infant health but have other limitations. We consider patterns of results across the different measures of poor infant health, and in auxiliary analyses, explore the potential for bias in our estimates.

The first measure of poor infant health—very severe infant health condition—coded from the medical records and one-year maternal reports of child disability, is whether the infant had a major abnormal health condition at birth. The coding of abnormal conditions was conducted by an outside pediatric consultant who was directed to identify cases that were severe; unlikely caused by prenatal behavior; had a poor long-term prognosis; and, in the case of one-year maternal reports, were likely present at birth. Our goal was to capture conditions that are, for the most part, random (e.g., Down syndrome, congenital heart malformations), given that the pregnancy resulted in a live birth. This measure most closely matches our criteria for an exogenous health shock. However, extreme conditions were rare: only 2% of the children in each of our analysis samples had a very severe infant health condition as we have defined it (Table 1).

The second measure of poor infant health—severe infant health condition—is very similar to the first measure (very severe infant health condition). However, it also includes children with very low birth weight (<1,500 grams) but had no severe abnormal conditions.11 Very low birth weight is a strong risk factor for a number of serious and long-term child health conditions (Reichman 2005). In both analysis samples, 3% had a severe infant health condition as we have defined it (Table 1). The advantage of this measure is that it yields a larger number of cases coded as having poor infant health, all of whom have a high probability of having long-term health problems. The disadvantage is that it is less random because very low birth weight can be related to socioeconomic status and prenatal behaviors (Reichman 2005).

The third is a direct but broad measure of poor infant health: whether the infant had a severe or moderately severe abnormal condition (see footnote 11). This measure, which does not take into consideration birth weight, includes conditions that are most likely unrelated to prenatal behavior but that may or may not have poor long-term prognoses. We call this measure any infant health condition. Again, the coding was conducted by an outside pediatric consultant. One-fifth of the children in each analysis sample were coded as having any infant health condition as we have defined it (Table 1).12

For our fourth measure of poor infant health, we use low birth weight (<2,500 grams). This measure is readily obtained from maternal reports or medical records, but it is not very specific because few children with moderately low birth weight (the majority of those with low birth weight)—namely, those weighing between 1,500 and 2,500 grams—have severe health problems (Reichman 2005). A distinct disadvantage is that low birth weight is known to be associated with poverty and prenatal behavior. The value of using low birth weight as a measure of poor infant health is that it is a widely used marker and is comparable across studies. In both analysis samples, 9% of the infants had low birth weight (Table 1). Because this measure is strongly associated with poverty and prenatal behavior, it is likely endogenous. We include it strictly for comparison purposes.

Covariates

We include variables that represent the arguments in Eq. 1. First, we include the father’s criminal activity prior to the birth, which is related to his human and social capital. Incarceration tends to both improve illegal job skills and erode legal job skills (Freeman 1996), and having a criminal record may limit legal employment options. Raphael (2007) found that previously incarcerated young men are less likely than those who have not been incarcerated to make successful transitions into adulthood; they are also more likely to live with their parents as adults, less likely to marry, less likely to be employed, and more likely to have low hourly earnings. Controlling for fathers’ pre-birth conviction (in models of conviction) and incarceration (in models of incarceration) allows us to capture these different starting points and to assess changes in criminal activity as a result of the infant health shock.

Second, we include demographic and economic characteristics of the father that are related to his human and social capital (all measured at baseline). We include the father’s age (in five-year intervals), education (high school graduate and any college, compared with less than high school), current employment, and ever having served in the military, which are measures of human capital. We include whether the father lived with both of his parents at age 15, which is a proxy for social ties and disadvantage during childhood, and the percentage of households in the father’s census tract with income under the poverty line, which is a proxy for household income as well as for neighborhood conditions. We also include the father’s nativity (foreign-born versus U.S.-born) and race/ethnicity (non-Hispanic black, Hispanic, and other, compared with non-Hispanic white).

Third, we include measures of the parents’ baseline relationship status (whether they were married, cohabiting, or living apart) and other measures related to family ties or commitment (parents knew each other at least twelve months at the time of the birth, whether the father visited the mother and baby in the hospital, and previous fertility of each parent). About one-quarter of the parents were married at the time of the birth (by design). Of those who were unmarried, more than one-half were cohabiting. We include several characteristics of the mother (age, race/ethnicity, education, and employment status),13 which may be related to the father’s human and social capital. We also include the focal child’s gender because boys are more likely than girls to be in poor health (Verbrugge 1982) and to have involved fathers (Dahl and Moretti 2008).

As indicated earlier, prior research has found that criminal justice sanctions, labor market characteristics, and other macro-level variables have effects on crime. In order to control for geographical variations in the probability of arrest, conviction, and incarceration, as well as for other city or state characteristics that may affect criminal behavior and possibly infant health, all models include indicators for the city in which the birth occurred (not shown in tables).

Multivariate Results

Primary Analyses

Probit estimates of the effects of each of the four measures of poor infant health on the two different measures of the father’s post-birth criminal activity are presented in Table 2 (each cell represents a separate probit model). In each cell, the probit coefficient appears on top, the marginal effect is in brackets, and the standard error of the marginal effect, corrected for city clustering of observations using the Huber-White method, is in parentheses.

Three of the measures of poor infant health (very severe infant health condition, severe infant health condition, and low birth weight) significantly increase the likelihood that the father is convicted after the birth of the child, by 2 to 8 percentage points with controls for pre-birth conviction, the covariates listed in Table 1, and city indicators. For incarceration (holding constant pre-birth incarceration), the estimates have largely the same pattern but are smaller in magnitude than those for conviction and are statistically significant only for any infant health condition and low birth weight (at the 10% level). It is noteworthy that the magnitude is greatest for our most stringent measure of poor infant health shocks (very severe infant health condition) and lowest for our broadest measure (any infant health condition). It is also noteworthy that the marginal effect of low birth weight (a less precise measure of poor long-term health than the measures of severe conditions and also the most likely of the four measures to be endogenous) is similar in magnitude to that of any infant health condition (a less precise measure of poor long-term health than the measures of severe conditions). Finally, it is validating that the directional effects are similar for the two different outcomes and that the effects are greater in magnitude for conviction than for incarceration. As discussed earlier, this was to be expected, given that not all convictions result in incarceration and that incarceration takes place later in the crime-conviction-incarceration process. Overall, these results suggest that having a seriously unhealthy infant increases the likelihood that the father will commit crime during the child’s early formative years, and that the more severe the health condition, the greater the effect.

Table 3 presents the multivariate results for the first two measures of poor infant health from Table 2. Although the covariate estimates vary somewhat depending on the measure of criminal activity, they paint a remarkably consistent picture. As expected, having been convicted or incarcerated prior to the child’s birth is a strong predictor of being convicted (first two columns) or incarcerated (second two columns) after the birth of the child, as are sociodemographic characteristics and measures of the father’s human capital. Fathers who graduated from high school or attended at least some college are less likely than those who did not graduate from high school to be convicted of or incarcerated for a crime after the child’s birth; this result is consistent with findings by Lochner and Moretti (2004) of strong effects of high school completion on incarceration when the potential endogeneity of education is controlled for. In addition, fathers who were employed at the time of the child’s birth are less likely than those who were not employed to later be convicted of or incarcerated for a crime.

As expected, fathers who were not married to the child’s mother at the time of the child’s birth were more likely than those who were married to the child’s mother to be convicted or incarcerated after the birth of the child. However, unmarried fathers who were cohabiting with the mother at baseline were as likely as those not living with the mother at baseline to later be convicted or incarcerated, which is consistent with findings of Duncan et al. (2006) of stronger negative associations between marriage and substance use than between cohabitation and substance use.

The likelihood that the father was convicted or incarcerated after the birth is negatively associated with his age. Although the coefficients are not always statistically significant, the pattern is consistent with the life cycle of criminal activity discussed earlier. In this sample of urban men, most of whom fathered children out of wedlock, we find no associations between race/ethnicity and crime in our multivariate models.14 Immigrant status is strongly and negatively related to incarceration; men who are immigrants are 5 percentage points less likely than native-born fathers to be incarcerated. These results are consistent with those of Butcher and Piehl (1998, 2007), who found that recent immigrants are a select group with lower levels of criminal activity than their nonimmigrant peers. Living in a poorer census tract at baseline significantly increases the likelihood of future conviction. We do not find a significant association between neighborhood poverty and incarceration.

Altogether, our results are strong and robust. Holding constant past criminal activity and numerous covariates, the shock of having an infant in poor health increases the likelihood that the father will become convicted of a crime. The estimates are in the expected pattern across our different measures of poor infant health and across two commonly used measures of involvement with the criminal justice system.

Specification Checks

Our analyses rely on the assumptions that poor infant health causes paternal criminal behavior (not the other way around) and that our measures of poor infant health are random shocks. For three of our measures of poor infant health—very severe infant health condition, severe infant health condition, and any infant health condition—we used an outside pediatric consultant to identify conditions that the medical community believe to be unrelated to prenatal behavior. Severe infant health condition was supplemented with cases of very low birth weight. We explore the possibility that despite this strategy, our rich set of controls, and the validating patterns across measures of infant health and outcomes, there remain unobserved factors underlying the associations between poor infant health and fathers’ criminal behavior. Because we cannot formally test for endogeneity, we explore potential bias in our estimates by conducting and evaluating the results from a series of supplemental analyses (results not shown, but available upon request).

One potential explanation of why poor infant health is related to post-birth criminal activity is that the causality runs in the other direction. We know that pre-birth criminal activity is highly correlated with post-birth criminal activity. It is possible that the father’s pre-birth criminal behavior caused the mother to experience stress while pregnant and that the stress adversely affected the infant’s health. To test this potential alternative explanation, we estimated poor infant health as a function of pre-birth criminal behavior and the other covariates, using each of the measures of (pre-birth) criminal activity (conviction or incarceration) and each of the four measures of poor infant health.15 These results indicate that the father’s pre-birth criminal activity is not associated with our coded measures of poor infant health. The association is, however, positive and significant for low birth weight and the father’s incarceration before the birth. Finding no association between pre-birth crime and infant health using our coded measures of infant health shocks but finding a significant positive association between pre-birth crime and low birth weight (a measure we assumed from the outset was endogenous) suggests that we have been successful in characterizing infant health shocks with our coded measures of poor infant health.

Second, we explored the possibility that the observed associations between poor infant health and father’s criminal activity actually reflect attributes of the mother. To explore the potential role of the mother’s taste for risk, we estimated models equivalent to those in Table 3 that included measures of whether the mother smoked cigarettes, used illicit drugs, and drank alcohol during the pregnancy, as well as whether she initiated prenatal care after the first trimester.16 We found that including those prenatal behaviors slightly decreased the estimated effects of poor infant health on crime, but never significantly so (except, as consistent with our expectations and other specification checks, when predicting incarceration as a function of low birth weight); this finding suggests that the mother’s taste for risk is not a viable competing explanation for our findings.

Finally, the processing of arrestees through the criminal justice system takes time, and it is therefore possible that some fathers who were convicted or incarcerated after the child was born had actually committed the crime prior to the birth. According to the U.S. Department of Justice (2004), the median length of time between arrest and sentencing for those convicted of a felony in a state court in 2002 was 184 days.17 To account for this timing issue, we ran supplemental models for which we redefined post-birth criminal activity as conviction (incarceration) one or more years after the birth of the child. We used the one-year period because more than three-quarters of felony sentences occur within one year after arrest (U.S. Department of Justice 2004). The marginal effects of poor infant health on post-birth criminal activity in these models were stronger for both the very severe infant health conditions and severe infant health conditions than the marginal effects in Table 2. However, the difference was statistically significant only for very severe infant health condition when considering conviction and for severe infant health condition when considering incarceration. The marginal effects of any infant health condition and low birth weight were never significantly different from those in Table 2. Thus, it appears that timing lags from crime to conviction and incarceration are not confounding the estimates.

Interactive Effects

As discussed earlier, we expect that an adverse life shock would have a stronger negative effect on crime for individuals who are more at risk for engaging in criminal activities by virtue of lower initial (pre-baseline) levels of human and social capital. To test this hypothesis, we estimated models corresponding to those in Table 3 that interacted poor infant health with baseline measures of weak ties to the child’s mother (neither married nor cohabiting), weak labor market ties (not employed), living in a poor census tract (30% or more of families below poverty), or prior conviction or incarceration (results not shown, but are available upon request). As expected, the effects of poor infant health on paternal crime were uniformly greater for fathers with weaker ties.18 Notably and not surprisingly, the effects of poor infant health on crime were at least four times greater for fathers with a prior criminal record than for those without a prior criminal record. This result is consistent with the turning point hypothesis that strong social ties lead to desistance from crime.

Next, we investigated the extent to which the life shock appears to be a catalyst to the launching of a criminal career. We know from the literature on criminal careers that the likelihood of entering into crime is greater at younger ages (Piquero et al. 2003). Although the turning point literature has focused primarily on patterns of desistance, some individuals initiate criminal behavior during their adult years (Sampson and Laub 1990). To explore the possibility that an event that weakens social bonds can impel a father into crime and that this is more likely to occur for younger men rather than older men, we restricted the sample to fathers who had neither been convicted nor incarcerated before the birth of the child and estimated models that interacted poor infant health and young paternal age (younger than 25 years). We found that for older fathers, the effects of poor infant health on crime were small in magnitude and statistically insignificant, whereas for younger fathers, the effects were large in magnitude (point estimates ranging from 0.11 to 0.27) and statistically significant. These findings indicate that for young adult men, turning points can lead to entry into crime.

Potential Mechanisms

As discussed earlier, previous research has found evidence that poor infant health adversely affects a father’s community and social bonds as measured by his labor supply and his coresidence status with the mother of the child. As such, the effects of poor infant health on paternal crime may operate, at least to some extent, through those channels. To explore paternal labor supply and parental relationship status as potential mediators of the effects of poor infant health on crime, we estimated a set of supplementary models that, in addition to all the same covariates as the models in Table 3, include measures of paternal labor supply and/or relationship status. The former is measured as the number of weeks the father worked in the past year, and the latter is characterized by whether the parents lived together (were married or cohabiting). Because it was not possible to firmly establish the temporal ordering from poor infant health to the mediators to crime, we estimated some models with the mediators measured at one year and others with the mediators measured at three years. Another reason for considering the two different time points was that paternal incarceration could complicate analyses of these potential mediators: for example, if a father is incarcerated at the time of a given follow-up interview, he cannot be living with the mother. Estimates from the analyses of weeks worked and living with the mother are presented in Table 4. Each row represents a set of models designed to explore the role of the mediator(s) designated in the row heading. For each row and measure of poor infant health, the first column presents the estimated effects of poor infant health when the mediator(s) is (are) added to the model, and the second column presents the estimated effect of the mediator(s).

We find that the mediators, whether measured at one or three years, tend to be significant predictors of crime,19 always have coefficients with the expected sign, and usually reduce the magnitude of the marginal effects of poor infant health on crime. Including both mediators together, whether measured at one or three years, reduces the marginal effects of poor infant health on crime by about 20%, although the changes in magnitude from the corresponding models without the mediators (from Table 3) are not always statistically significant. That said, the consistent patterns for the same mediators measured at different time points and across different measures of poor infant health and outcomes provide some suggestive evidence that the effects of poor infant health on crime operate partly through the impact of poor infant health on paternal labor supply and parental relationship status. When we assessed the sensitivity of these estimates to alternative definitions of labor supply and relationship status (e.g., using employed at all in the past year and romantically involved at all at the time of a given follow-up interview), we found the same general patterns. Also, analyses using conviction between the one- and three-year follow-up interviews and incarceration between the one-year and three-year follow-up interviews as the outcomes also suggest that work and relationship status play mediating roles.

Finally, we explored paternal mental health, residential moves, and the mother’s reliance on cash assistance as potential mediators (results not shown). Mental health might affect an individual’s assessment of the relative costs and benefits of committing crime. Residential relocation, which can result from changes in relationships, work, or financial circumstances, may increase crime by weakening or severing the father’s bonds with friends, extended family, and community. In terms of cash assistance participation, past research indicating that poor infant health reduces the mother’s labor supply (Corman et al. 2005) and increases her reliance on welfare (Reichman et al. 2006) suggests that fathers of infants born in poor health may face increased child support obligations, which might make illegal work more attractive. We found that while all these factors measured at one or three years were strongly associated with crime in the expected directions, including them in the models rarely changed the estimated effects of poor infant health.

Overall, we find evidence that work and relationship status (particularly work) appear to explain at least part of the effects of poor infant health on paternal crime, but that the father’s mental health, the father’s residential relocation, and the mother’s reliance on cash assistance do not. However, because we are not able to firmly establish the temporal sequences between mediating factors and outcomes, these findings should be considered preliminary.

Conclusion

We found evidence that a specific life event affects the propensity of at-risk men to commit crime. Specifically, we found that the birth of an infant with a severe health problem increases the likelihood that the child’s father will become convicted of a crime and that work and parental relationships appear to mediate the effects to some extent. We studied a contemporary population-based sample from a birth cohort study of unmarried parents in large U.S. cities. These fathers tended to be poor and uneducated, and therefore were at high risk for committing crime. We know of no other population-based studies of effects of a life event on criminal activity of high-risk men. We used two different proxies for criminal behavior: conviction and incarceration, the latter of which we could cross-validate using paternal and maternal reports. Our measures of poor infant health were designed to characterize random shocks, and we conducted a number of specification checks to test our assumption of exogeneity.

The findings from this study provide evidence in support of the turning point hypothesis and complement recent studies exploiting natural experiments to estimate the effects of marriage on crime (Caceres-Delpiano and Giolito 2008; Edlund et al. 2007). The findings also inform the relatively recent literature on the effects of incarceration on child and family well-being (e.g., Geller et al. 2009; Wildeman 2009, 2010) by underscoring the potential importance of reverse pathways. Additionally, this research contributes to both a virtually non-existent literature on the effects of fatherhood on men’s social behavior (see Edin et al. 2004) and a small but growing literature on the effects of child health on family resources (Reichman et al. 2008).

We conclude with three caveats. First, the results may not be generalizable to other types of adverse life shocks. Second, the results pertain to men who are fathers and may not be generalizable to all men at risk for committing crime. Young adult men who transition to fatherhood may respond to life shocks differently than those who do not become fathers. Third, the men in the Fragile Families Study who were not in the sample because of attrition from the study are more socioeconomically disadvantaged than those in the analysis sample. Based on the results from our interaction models, this should lead to underestimates of the effects of poor infant health on crime.

Acknowledgments

This research was supported by Grants R01-HD-45630 and R01-HD-35301 from the National Institute of Child Health and Human Development. We are grateful for helpful input from Jerry Bentley, Naci Mocan, Alan Monheit, Melinda Pitts, Joy Schneer, and the participants at the economic seminar series at Lafayette College and the University of Medicine and Dentistry School of Public Health, and for valuable assistance from Magdalena Ostatkiewicz, Nicole Boynton, and Prisca Figaro.

Appendix: Coding of Measures of Poor Infant Health

The coding of abnormal conditions was designed to identify cases that were at least moderately severe, were not likely caused by prenatal behavior, had a poor long-term prognosis, and were present at birth. A pediatric consultant was directed to glean information from the medical records (augmented with one-year maternal reports) and to assign all infant conditions a number between 1 and 16 according to the grid shown in Table 5. After giving the consultant the grid and clear instructions, the authors had no further input into how particular conditions were coded. If a child had multiple conditions, each condition was assigned a separate number. Very severe infant health condition was coded as 1 (yes) if the child had a health condition in cell 1. Examples of conditions in cell 1 are microcephalus, renal agenesis, total blindness, and Down syndrome.

Severe infant health condition was coded as 1 (yes) if the child had a condition in cell 1 or the child was very low birth weight (less than 1,500 grams).Any infant health condition was coded as 1 (yes) if the child had a condition in either cell 1 or cell 2. Examples of conditions in cell 2, which are considered random at birth but may or may not have long-term health consequences, are malformed genitalia, hydrocephalus, cleft palate, shoulder dystocia, pneumomediastinum, and webbed fingers or toes.

Notes

1

Two exceptions are studies by Sampson and Laub (1996), which investigated effects of military service on social bonds among servicemen during World War II, and Edin et al. (2004), which examined the effects of fatherhood on crime in a qualitative study of 200 low-income noncustodial fathers.

2

The oversampling of nonmarital births solely in urban areas resulted in a sample that, generally, is more disadvantaged than the general population. See Reichman et al. (2001) for a description of the research design.

3

Access to the medical records reflected, to a large extent, administrative decisions of the different hospitals rather than decisions on the part of individual respondents to have their records reviewed.

4

Of course, individuals can be convicted of crimes that they did not commit, or can be arrested and incarcerated before trial and then ultimately acquitted or have the charges dropped. Such cases would result in false positives for convictions and incarcerations, respectively.

5

Although the turning point literature focuses on the long-term trajectory of criminal activity, the effects of life events on crime appear to occur rather quickly. For example, Laub et al. (1998) found that by three to four years of married life, arrest rates are 52.5% lower for men in strong marriages than for unmarried men, and that the difference increases only minimally over the next two years (to 53.8%) and then tapers off quickly. Thus, our three-year time horizon seems reasonable.

6

Compared with the men in our sample, the men not in the sample were significantly less likely to have been married to their child’s mother, less likely to have been in a relationship with the mother for at least 12 months, less likely to have visited the mother and child in the hospital, less likely to have lived with both parents at age 15, and more likely to be non-Hispanic black.

7

Small cell sizes and incomplete data on types of crime preclude separate analyses by type of crime.

8

The following questions were used from the fathers’ follow-up surveys to determine whether the father had ever been convicted after the birth of the child and whether the father had ever been convicted before the birth of the child: (1) “Have you ever been convicted of any charges?” (2) “How old were you (the first time/when) this happened?” (3) “When was your most recent conviction?”

9

The incarceration questions asked of fathers in the follow-up surveys were similar to those for conviction (see footnote 8). The questions asked of mothers about the fathers’ incarceration history were similar in structure to those asked of fathers. We used fathers’ reports when they were available and mothers’ reports otherwise. Fathers’ reports were used for 2,469 of the 2,677 cases in the analysis sample for incarceration. There were only 18 cases in which the father reported that he had never been incarcerated but the mother gave conflicting information, and the estimates were insensitive to the coding of those cases. The high level of agreement between the father and mother reports of incarceration and the insensitivity of our estimates are validating of our incarceration measure. Additional details about the construction of both the incarceration and conviction measures are available upon request.

10

It is possible that judges are less willing to incarcerate convicted fathers with unhealthy children than to incarcerate convicted fathers with healthy children. We know of no evidence on this, but if this is the case, it would result in underestimated effects of poor infant health on crime using the incarceration measure. This potential scenario underscores the value in considering conviction as an alternative proxy for crime.

11

All but two of the infants with very low birth weight also had moderately severe infant health conditions, which are defined as conditions not considered to be related to maternal behavior that may or may not have long-term health consequences.

12

Because there is neither a standard for measurement nor a consistent reporting of child disability (Reichman et al. 2008), it is difficult to provide a good national comparison for our figures. Our rates of poor infant health are consistent with the range of 6% to 18% of children in the United States that have special health care needs, as reported by Stein (2005). It is not unexpected that our strictest measure is lower than this range, since that measure includes only very serious conditions and excludes conditions that are known to be related to maternal behaviors. Likewise, it is not unexpected that our broadest measure is slightly higher than the upper-bound estimate, since it is defined to include conditions that are not necessarily disabling in the long run.

13

Because of high collinearity between mothers’ and fathers’ ages, race/ethnicity, and education, these measures are expressed as differences from the father.

14

Considering the characteristics of our sample, our unadjusted race-specific rates of conviction and incarceration are consistent with racial differences in crime observed at the national level. For example, in 2005, about 28% of all arrests in the United States were to blacks, who made up about 14% of the U.S. population (U.S. Department of Justice 2006). In our sample, 24% of non-Hispanic black fathers and 18% of non-Hispanic white fathers had been convicted of a crime before the focal child’s birth, and 18% of non-Hispanic black fathers and 9% of non-Hispanic white fathers had been incarcerated before the focal child’s birth.

15

We also ran a set of “falsification tests” that estimated pre-birth criminal activity as a function of poor infant health and the other covariates. The logic was that a shock that takes place at the time of the birth cannot possibly affect the father’s pre-birth criminal activity, and finding significant associations would indicate spurious correlation. These models are essentially equivalent to the models of poor infant health as a function of pre-birth criminal behavior, except that the dependent and independent variables were reversed. As expected, we found that poor infant health was an insignificant predictor of pre-birth crime in all cases except when using low birth weight and predicting incarceration.

16

See Reichman et al. (2009) for more detail on the construction of the smoking, drug use, and prenatal care measures, which incorporated medical records and survey data. The alcohol measure was constructed using the same methodology as for the smoking and drug use variables.

17

We do not know whether the men in our sample were convicted (or incarcerated) for a misdemeanor or for a felony. We would expect the processing time to be shorter for misdemeanors.

18

The interactions were statistically significant at conventional levels in half of the cases. They were most consistently significant (across measures of poor infant health and outcomes) for unemployed and having prior criminal activity.

19

Exceptions are for living with the mother at one year, for which p values range from .09 to .18.

References

Becker, G. (
1968
).
Crime and punishment: An economic approach
.
Journal of Political Economy
,
76
,
169
217
. 10.1086/259394
Butcher, K. F., & Piehl, A. M. (
1998
).
Recent immigrants: Unexpected implications for crime and incarceration
.
Industrial and Labor Relations Review
,
51
,
654
679
. 10.2307/2525013
Butcher, K. F., & Piehl, A. M. (2007). Why are immigrants’ incarceration rates so low? Evidence on selective immigration, deterrence, and deportation (NBER Working Paper No. 13229). Cambridge, MA: National Bureau of Economic Research.
Caceres-Delpiano, J., & Giolito, E. (2008). The impact of unilateral divorce on crime (IZA Discussion Paper No. 3380). Bonn, Germany: Institute for the Study of Labor.
Capaldi, D. M., Kim, H. K., & Owen, L. D. (
2008
).
Romantic partners’ influence on men’s likelihood of arrest in early adulthood
.
Criminology
,
46
,
267
299
. 10.1111/j.1745-9125.2008.00110.x
Corman, H., Noonan, K., & Reichman, N. E. (
2005
).
Mothers’ labor supply in fragile families: The role of child health
.
Eastern Economic Journal
,
31
,
601
616
.
Dahl, G. B., & Moretti, E. (
2008
).
The demand for sons
.
Review of Economic Studies
,
75
,
1085
1120
. 10.1111/j.1467-937X.2008.00514.x
Duncan, G. J., Wilkerson, B., & England, P. (
2006
).
Cleaning up their act: The effects of marriage and cohabitation on licit and illicit drug use
.
Demography
,
43
,
691
710
. 10.1353/dem.2006.0032
Edin, K., & Reed, J. M. (
2005
).
Why don’t they just get married? Barriers to marriage among the disadvantaged
.
The Future of Children
,
15
,
117
137
. 10.1353/foc.2005.0017
Edin, K., Nelson, T. J., & Paranal, R. (
2004
).
Fatherhood and incarceration as potential turning points in the criminal careers of unskilled men
. In Pattillo, M., Weiman, D., & Western, B. (Eds.),
Imprisoning America: The social effects of mass incarceration
(pp.
46
75
).
New York
:
Russell Sage Foundation
.
Edlund, L., Li, H., Yi, J., & Zhang, J. (2007). Sex ratios and crime: Evidence from China’s one-child policy (IZA Discussion Paper No. 3214). Bonn, Germany: Institute for the Study of Labor.
Ehrlich, I. (
1973
).
Participation in illegitimate activities: A theoretical and empirical investigation
.
Journal of Political Economy
,
81
,
521
565
. 10.1086/260058
Freeman, R. B. (
1996
).
Why do so many young American men commit crimes and what might we do about it?
.
Journal of Economic Perspectives
,
10
,
461
488
.
Geller, A., Garfinkel, I., Cooper, C., & Mincy, R. (
2009
).
Parental incarceration and child wellbeing: Implications for urban families
.
Social Science Quarterly
,
90
,
1186
1202
. 10.1111/j.1540-6237.2009.00653.x
Gould, E. D., Weinberg, B. A., & Mustard, D. B. (
2002
).
Crime rates and local labor market opportunities in the United States 1979–1997
.
Review of Economics and Statistics
,
84
,
45
61
. 10.1162/003465302317331919
Grogger, J. (
1998
).
Market wages and youth crime
.
Journal of Labor Economics
,
16
,
756
791
. 10.1086/209905
Horney, J., Osgood, W., & Marshall, I. H. (
1995
).
Criminal careers in the short-term: Intra-individual variability in crime and its relation to local life circumstances
.
American Sociological Review
,
60
,
655
673
. 10.2307/2096316
Kerner, H-J (
2005
).
The complex dynamics of the onset, the development, and the termination of a criminal career: Lessons on repeat offenders to be drawn from recent longitudinal studies in criminology
.
The Annals of the American Academy of Political and Social Science
,
602
,
259
279
. 10.1177/0002716205281245
King, R. D., Massoglia, M., & MacMillan, R. (
2007
).
The context of marriage and crime: Gender, the propensity to marry, and offending in early adulthood
.
Criminology
,
45
,
33
65
. 10.1111/j.1745-9125.2007.00071.x
Laub, J. H., & Sampson, R. J. (
1993
).
Turning points in the life course: Why change matters to the study of crime
.
Criminology
,
31
,
301
325
. 10.1111/j.1745-9125.1993.tb01132.x
Laub, J. H., Nagin, D. S., & Sampson, R. J. (
1998
).
Trajectories of change in criminal offending: Good marriages and the desistance process
.
American Sociological Review
,
63
,
225
238
. 10.2307/2657324
Lochner, L., & Moretti, E. (
2004
).
The effect of education on crime: Evidence from prison inmates, arrests, and self-reports
.
American Economic Review
,
94
,
155
189
. 10.1257/000282804322970751
McLanahan, S., Garfinkel, I., Reichman, N. E., Teitler, J. O., Carlson, M., & Norland Audigier, C. (2003). The fragile families and child wellbeing study: Baseline national report (revised March 2003). Princeton, NJ: Center for Research on Child Wellbeing. Retrieved from http://www.fragilefamilies.princeton.edu/documents/nationalreport.pdf
Mocan, N., & Rees, D. I. (
2005
).
Economic conditions, deterrence and juvenile crime: Evidence from micro data
.
American Law and Economics Review
,
7
,
319
349
. 10.1093/aler/ahi011
National Center for Health Statistics. (2011). Health, United States, 2010 with special feature on death and dying. Hyattsville, MD: U.S. Government Printing Office. Retrieved from http://www.cdc.gov/nchs/hus.htm
Noonan, K., Reichman, N. E., & Corman, H. (
2005
).
New fathers’ labor supply: Does child health matter?
.
Social Science Quarterly
,
86
,
1399
1417
. 10.1111/j.0038-4941.2005.00352.x
Piquero, A. R., Farrington, D. P., & Blumstein, A. (
2003
).
The criminal career paradigm
.
Crime and Justice
,
30
,
359
506
.
Raphael, S. (
2007
).
Early incarceration spells and the transition to adulthood
. In Danziger, S., & Rouse, C. E. (Eds.),
The price of independence: The economics of early adulthood
(pp.
278
306
).
New York
:
Russell Sage Foundation
.
Reichman, N. E. (
2005
).
Low birth weight and school readiness
.
The Future of Children
,
15
,
91
116
. 10.1353/foc.2005.0008
Reichman, N. E., Teitler, J. O., Garfinkel, I., & McLanahan, S. (
2001
).
Fragile families: Sample and design
.
Children and Youth Services Review
,
23
,
303
326
. 10.1016/S0190-7409(01)00141-4
Reichman, N. E., Corman, H., & Noonan, K. (
2004
).
Effects of child health on parents’ relationship status
.
Demography
,
41
,
569
584
. 10.1353/dem.2004.0026
Reichman, N. E., Corman, H., & Noonan, K. (
2006
).
Effects of child health on sources of public support
.
Southern Economic Journal
,
73
,
136
156
. 10.2307/20111879
Reichman, N. E., Corman, H., & Noonan, K. (
2008
).
Impact of child disability on the family
.
Maternal and Child Health Journal
,
12
,
679
683
. 10.1007/s10995-007-0307-z
Reichman, N. E., Corman, H., Noonan, K., & Dave, D. M. (
2009
).
Infant health production functions: What a difference the data make
.
Health Economics
,
18
,
761
782
. 10.1002/hec.1402
Saluter, A. F. (1994). Marital status and living arrangements: March 1994 (Current Population Reports, Series P20-484). Washington, DC: U.S. Government Printing Office. Retrieved from http://www.census.gov/prod/1/pop/p20-484.pdf
Sampson, R. J., & Laub, J. H. (
1990
).
Crime and deviance over the life course: The salience of adult social bonds
.
American Sociological Review
,
55
,
609
627
. 10.2307/2095859
Sampson, R. J., & Laub, J. H. (
1996
).
Socioeconomic achievement in the life course of disadvantaged men: Military service as a turning point, circa 1940–1965
.
American Sociological Review
,
61
,
347
367
. 10.2307/2096353
Sampson, R. J., Laub, J. H., & Wimer, C. (
2006
).
Does marriage reduce crime? A counterfactual approach to within-individual causal effects
.
Criminology
,
44
,
465
508
. 10.1111/j.1745-9125.2006.00055.x
Sickles, R. C., & Williams, J. (
2008
).
Turning from crime: A dynamic perspective
.
Journal of Econometrics
,
145
,
158
173
. 10.1016/j.jeconom.2008.05.014
Stein, R. E. K. (
2005
).
Trends in disability in early life
. In Fields, M. J., Jerre, A. M., & Marin, L. (Eds.),
Workshop on disability in America: A new look
(pp.
143
156
).
Washington, DC
:
The National Academies
.
The Cost of Crime: Understanding the Financial Impact of Criminal Activity: Testimony before the U.S. Senate Judiciary Committee, Bureau of Justice Statistics, 109th Cong. 4 (2006). (testimony of Jeffery Sedgwick).
U.S. Census Bureau. (2009). America’s families and living arrangements: 2008. Retrieved from http://www.census.gov/population/www/socdemo/hh-fam/cps2008.html
Felony sentences in state courts, 2002 (Bulletin NCJ 206916)
. (
2004
).
Washington, DC
:
Bureau of Justice Statistics
.
Verbrugge, L. M. (
1982
).
Sex differentials in health
.
Prevention
,
97
,
417
437
.
Wildeman, C. (
2009
).
Parental imprisonment, the prison boom, and the concentration of childhood disadvantage
.
Demography
,
46
,
265
280
. 10.1353/dem.0.0052
Wildeman, C. (
2010
).
Paternal incarceration and children’s physically aggressive behaviors: Evidence from the fragile families and child wellbeing study
.
Social Forces
,
89
,
285
309
. 10.1353/sof.2010.0055