Children Who Prosper in Unfavorable Environments: The Relationship to Social Capital
Objective. Social capital describes the benefits that are derived from personal social relationships (within families and communities) and social affiliations. This investigation examined the extent to which social capital is associated with positive developmental and behavioral outcomes in high-risk preschool children.
Design. A cross-sectional case-control analysis of young children “doing well” and “not doing well” at baseline in four coordinated longitudinal studies.
Participants. A total of 667 2- to 5-year-old children (mean age, 4.4 years) and their maternal caregivers who are participating in the Longitudinal Studies of Child Abuse and Neglect Consortium. At recruitment, all children were characterized by unfavorable social or economic circumstances that contributed to the identification of the children as high risk.
Measures. Social capital was defined as benefits that accrue from social relationships within communities and families. A social capital index was created by assigning one point to each of the following indicators: 1) two parents or parent-figures in the home; 2) social support of the maternal caregiver; 3) no more than two children in the family; 4) neighborhood support; and 5) regular church attendance. Outcomes were measured with the Child Behavior Checklist, a widely used measure of behavioral/emotional problems, and with the Battelle Developmental Inventory Screening Test, a standardized test that identifies developmental deficits. Children were classified as doing well if their scores on these instruments indicated neither behavioral nor developmental problems.
Results. Only 13% of the children were classified as doing well. The individual indicators that best discriminated between levels of child functioning were the most direct measures of social capital—church affiliation, perception of personal social support, and support within the neighborhood. The social capital index was strongly associated with child well-being, more so than any single indicator. The presence of any social capital indicator increased the odds of doing well by 29%; adding any two increased the odds of doing well by 66%.
Conclusions. Our findings suggest that social capital may have an impact on children's well-being as early as the preschool years. In these years it seems to be the parents' social capital that confers benefits on their offspring, just as children benefit from their parents' financial and human capital. Social capital may be most crucial for families who have fewer financial and educational resources. Our findings suggest that those interested in the healthy development of children, particularly children most at risk for poor developmental outcomes, must search for new and creative ways of supporting interpersonal relationships and strengthening the communities in which families carry out the daily activities of their lives.
- LONGSCAN =
- Longitudinal Studies of Child Abuse and Neglect •
- BDST =
- Battelle Developmental Inventory Screening Test •
- CBCL =
- Child Behavior Checklist •
- AAPI =
- Adult-Adolescent Parenting Inventory •
- CES-D =
- Center for Epidemiologic Studies Depression Scale •
- OR =
- odds ratio •
- CI =
- confidence interval
As a society, we are becoming increasingly concerned about the demise of close-knit neighborhoods and extended families. In 1988, James Coleman1 introduced the concept of social capital and the hypothesis that the benefits accrued from social connectedness in communities and within families impact the development and well-being of children. He defined social capital as those aspects of the social structure—personal relations and networks of relations—that facilitate actions within the structure. The other forms of capital, financial capital (money) and human capital (education), are widely recognized as resources for individual development and productivity; social capital refers to features in the social organization, such as social networks, expectations, and trust, that facilitate coordination and cooperation for mutual benefit. Social capital is derived from interpersonal relationships and an array of obligations, expectations, information channels, and norms within families and communities. It is conceptualized as a resource, like other forms of capital, that can be drawn on or accessed as needed.
Coleman1 used data from a national survey to examine social capital's impact on the formation of human capital, ie, high school completion. He examined five variables he believed either confer, or serve as proxies of, social capital: the presence of both parents in the household (more parental resources to invest in child), presence of one versus four siblings (fewer children represent a greater concentration of parental attention), fewer changes of school since fifth grade (social relations are disrupted with each move), regular attendance at religious services (organizational involvement is considered an important component of social capital and religious affiliation is the most common group membership among Americans2), and mother's high expectations for a child's educational attainment (reflecting family norms and parental investment in the child). These variables distinguished between adolescents who stayed in school and those who dropped out, with an additive effect noted when variables were examined in combination.
Since Coleman's1 seminal study, a few other investigators have begun to explore the impact of social capital on child well-being. In these studies, social capital—defined variably as a combination of factors including residential stability, social networks, family relationships, and community and organizational involvement—has been associated again with lower school drop-out rates3 and with decreased child behavior problems.4 No existing data collection effort, including Coleman's, has begun as a study of social capital. Measurements of social capital have been post-hoc and highly varied in content.
Early findings of significant associations between diminished social capital and both high school drop-out and child behavior problems suggest that social capital may have an impact on child development. We speculate that social capital may be a crucial mediator in the relationship between adverse family conditions (such as poverty or domestic violence) and the development of children living in these conditions. A seminal article by Sameroff et al,5published in this journal in 1987, described how a cumulative environmental risk index predicted intelligence quotient scores at age 4 better than any single risk factor. Building on this work that highlighted the importance of the child's environment by examining multiple environmental indicators in concert, we implemented a reciprocal study to determine the extent to which an accumulation of social capital might exert a protective influence on children known to be at risk.
This study examines the cross-sectional relationship between an index of social capital and the developmental and behavioral well-being of young at-risk children through a case-control analysis of the Longitudinal Studies of Child Abuse and Neglect (LONGSCAN) baseline data.6 Extensive data are available on these children including five variables that parallel Coleman's1indicators of family and community social capital: 1) the presence of two parents or parent-figures in the household, 2) perception of social support by the child's maternal caregiver, 3) relatively fewer children in the household claiming family resources, 4) neighborhood support, and 5) regular church attendance by the family. This article examines whether the cumulative impact of social capital indicators is greater than the impact of each individual variable and tests the hypothesis that social capital is related to the developmental and behavioral well-being of high-risk preschool children.
The hypotheses were examined using previously-collected baseline data from a consortium of ongoing longitudinal studies of child maltreatment entitled LONGSCAN.6 Briefly, LONGSCAN is a coordinated set of five prospective studies of the causes, correlates, and consequences of child abuse and neglect. The studies use the same measures, data collection and handling procedures, and data entry system; they differ systematically by sampling children who: 1) are at-risk for maltreatment (North Carolina and Baltimore), 2) have been reported for maltreatment (Seattle), or 3) have been identified as maltreated and involved in remedial interventions by social service or other treatment agencies (Chicago and San Diego). The current article involves a cross-sectional analysis of baseline data from four of the LONGSCAN studies. These data include information on behavioral and developmental outcomes and on family and community characteristics, thus permitting the case-control analyses required for this study of social capital and child well-being. (The children from the fifth study site, Chicago, were infants at the time of the baseline interview and thus too young for the behavior problems assessment used in this analysis. Data collected at face-to-face interviews other than the baseline were not used, as data collection was still on-going at the time of the analysis reported here.)
The subjects for this investigation consisted of all LONGSCAN study children who were between the ages of 2 to 5 years at the time of the baseline interview: North Carolina (n = 141), San Diego (n = 146), Baltimore (n = 196), and Seattle (n = 184). Although the cohort samples were derived in diverse ways, all shared unfavorable environmental factors contributing to identification of these children as at-risk for child abuse or neglect.6In a North Carolina study that preceded LONGSCAN, consecutive infants born at participating hospitals were identified as at-risk because of factors such as low birth weight, gestational age <36 weeks, medical problems at birth, mother age 17 or less, parent with alcohol or substance abuse problem, no prenatal care, parent with incapacitating medical/mental handicap, or single parent with no family support. At 4 years of age, children from this sample who had been reported for maltreatment were enrolled in LONGSCAN along with a 2:1 comparison group of unreported children from the original cohort (matched for gender, income, race, and sex). The San Diego cohort is comprised of maltreated children placed in foster care in the first 42 months of life because of child maltreatment. At age 4, more than one-half of these children had returned to their original homes or moved to the home of a family member. The Baltimore sample was recruited from inner-city pediatric clinics serving low income families and consists of three groups: 1) children diagnosed as failing-to-thrive in the first 2 years of life, 2) children with mothers who were at-risk for human immunodeficiency virus, and 3) a comparison group without overt risk factors other than poverty. In the Seattle study, the sample consists of children, less than 5 years of age, reported to social services for maltreatment and classified as at moderate risk for future maltreatment on a state risk assessment instrument.6
Common measures and procedures were developed by the LONGSCAN consortium to address multiple questions related to risk and protective factors for child maltreatment and subsequent outcomes. After local Institutional Review Board approval, a set of common measures and procedures were implemented across all studies. Mothers, or other primary maternal caregivers, were asked to participate in a 2-hour face-to-face interview comprised of both standardized and project-developed measures. In addition, each child was administered a developmental screening test. The measures used are described in detail below. Data were entered locally using a common data entry system and were processed and analyzed at the LONGSCAN Coordinating Center at the University of North Carolina. A random 10% of the interviews were reentered to verify data entry procedures and coding. Cross-checks for invalid or incompatible responses were also conducted.
Definitions and Measures
Social capital was broadly defined as benefits that accrue from social relationships in communities and families. Five indicators of social capital were each identified and assessed as present or absent for each child's family. An index of social capital was created by assigning 1 point for the presence of each of the social capital indicators that follow, such that an individual's Social Capital Index score could range from 0 to 5.
Like Coleman,1 we considered the presence of two parents residing within the home as one indicator of social capital. If the maternal caregiver described a spouse or partner residing in the home, then two-parent family was classified as present.
Social support for the primary maternal caregiver was measured using the Affective and Confidante subscales of the Duke-UNC Functional Social Support Questionnaire.7 These scales are comprised of seven items describing access to social support with five response options ranging from 1 (“I get much less than I would like”) to 5 (“I get as much as I would like”). A sample item is “ … chances to talk to someone I trust about personal and family problems.” Social support was classified as present if the mean value of seven items was 4 (“ … almost as much as I would like”) or greater.
The presence of no more than two children in the home was scored as an indicator of social capital.
Neighborhood support was based on responses to three items from a nine-item project developed neighborhood scale. If the maternal respondent gave at least a somewhat affirmative response to questions about the presence of people in the neighborhood who help each other out, watch out for each other's children, and who can be counted on, she was classified as having neighborhood support.
If the maternal respondent indicated that she attended church or religious services at least several times per month during the past year, she was classified as having regular church involvement.
Child well-being was operationalized as obtaining developmental and behavioral scores well within normal limits based upon standardized measures. We assessed developmental outcomes with the Battelle Developmental Inventory Screening Test (BDST),8 a standardized test of developmental skills that has been shown to reliably identify children with developmental deficits. Behavioral outcomes were examined using the Internalizing, Externalizing, and Total Problems scores of the Child Behavior Checklist (CBCL),9 a widely used measure of behavioral/emotional problems with established reliability and predictive validity. Children were classified as “doing well” if their scores on these instruments indicated neither behavioral nor developmental problems. To be functioning well developmentally, a child needed a BDST Total score of 0 (within one standard deviation of the population mean for age or in the top 84% of the national norms). To be classified as functioning well behaviorally, a child's CBCL T-scores on the Total, Externalizing, and Internalizing Problems dimensions had to be ≤55, well below T = 60 which is the borderline cutpoint for the clinical range. The required scores on the two measures were generally comparable and set at a conservative level to ensure that children near the clinical borderlines on the two measures would not be classified as doing well.
Sociodemographic and Other Variables
The data on sociodemographic characteristics of the child and family were collected from the child's primary maternal caregiver at the time of the interview. These data on family income and maternal education describe the characteristics of the current caretaking family which is not necessarily the family of origin. Household income was assessed categorically with a response set of eleven levels beginning at less than $5000 per year and increasing by $5000 increments to the highest level (>$50 000/year). Care- giver depression was assessed using the Center for Epidemiologic Studies Depression Scale (CES-D)10 and parenting attitudes were assessed with the Adult-Adolescent Parenting Inventory (AAPI).11
We analyzed our cross-sectional data using a case-control approach. Cases were those study children classified as doing well based on the combined developmental and behavioral criteria; controls were the remaining study children. The intent of this definition of doing well was not to capture every child in the study that was functioning adequately but to define a group of children who were unequivocally doing well and examine factors that might account for that status. Other variables that have been theoretically or empirically linked to child well-being, such as sociodemographic factors (gender, race, age, mother's education, family income), parenting attitudes (as measured by the AAPI),11maternal depression (as measured by scores ≥16 on the CES-D ),10 and foster care placement, were also examined to assess the independent relationship of social capital to child well-being. We contrasted the children doing well with those doing less well: by each of the control variables (using the two-sample Wilcoxon rank sum test for continuous variables, the χ2 test of homogeneity for nondichotomous categorical variables, and Fisher's exact test for dichotomous categorical variables); by each of the five social capital indicators; and by the Social Capital Index dichotomized at ≥4 (using Fisher's exact test).
Using simple logistic regression models, we examined the relationship between the social capital index and each of the component criteria variables for defining “doing well” to determine if social capital might be differentially associated with child developmental outcomes as opposed to child behavioral outcomes. A second set of logistic regression models were constructed to look at the relationship between the social capital index and child functioning at each LONGSCAN study site.
Finally, the predictor variables that were univariately associated with overall child functioning or doing well, at an α level of .10, were considered simultaneously with maternal education and the Social Capital Index (levels 0 to 5) in a logistic regression model.12 Accordingly child age, gender, race, and parenting attitudes were not included in the final model. To account for systematic differences in sample selection among the study sites, indicator or dummy variables for the sites were included in the regression models. The indicator variable for a site had a value of 1 for a study child if that child belonged to that site, and 0 if not. Odds ratios (ORs) for child well-being, and their two-sided 95% confidence intervals (CI) were computed for all explanatory variables present in the model. Models including first-order interactions with site were implemented in a stepwise fashion but are not presented as the interaction terms were not statistically significant. Linear regression models were constructed to examine whether the decision to dichotomize each of the outcome variable components (CBCL scales and the BDST) and then pool them into a single binary outcome could have masked or exaggerated significant relationships. Statistical significance for all analyses was defined as P < .05. We used SAS 6.10 for OS/213 to perform the statistical computations.
There were 667 LONGSCAN baseline 2- to 5-year-old children included in this study. The mean age of the children was 4.4 years; there were 338 boys and 329 girls. Eighty-seven (13%) children were classified as doing well by the combined developmental and behavioral criteria. The BDST criterion alone excluded 488 of the 667 study children (73%) from the “doing well” category. The CBCL criteria alone excluded 385 children (58%) from the “doing well” category.
Table 1 displays the proportion of children doing well by site and demographic characteristics. Doing well did not vary significantly by child gender, child race, or maternal education. There were significant differences in child functioning by site, with fewer children doing well in San Diego and Seattle than in Baltimore and North Carolina (P = .02). Of the 56 children in foster care at the time of the assessment, only 3 met the criteria for doing well (P = .09).
Table 2 describes the demographic characteristics and maternal mental health and parenting scores for the cases (doing well) and controls (doing less well). Mothers of children doing less well were significantly more likely to be depressed (P = .02). Lower family income and foster care placement were marginally associated with the child function. As expected, the BDST Total score and the CBCL scores used for defining child well-being were all significantly lower among cases than controls.
As shown in Table 3, the unadjusted univariate ORs for the individual social capital indicators were all greater than one, meaning that the presence of each social capital indicator was synonymous with a higher odds of functioning well. Although regular church attendance and personal social support were the only indicators that were statistically significant, presence of neighborhood support exhibited a non-negligible relationship to doing well (P = .06). The OR of each of these three indicators of social capital was more than 1.50, implying at least a 50% increase in the odds of doing well in the presence of that indicator.
A dose-response relationship was noticed (see Fig1) when the percentage of children doing well was plotted against the social capital index value. The percentage of children doing well markedly increased from 15% for those with an index value of 4 to 39% for those with an index value of 5. When the social capital index was treated as a dichotomous variable (<4 vs ≥4), the OR for doing well was 1.78 (P = .04) for children with four or more social capital indicators present as compared with those with fewer than four social capital indicators. The dichotomized social capital index was more strongly associated with child functioning than any of the individual components of the index. When examined as an ordinal scale, with levels 0 to 5, the distribution of social capital was also significantly different between those children doing well and those doing less well (χ2 test of homogeneity statistic = 15.5; 5 df; P = .01).
The association between the social capital index (levels zero to five) and each of the four component criteria variables for defining doing well was tested separately in simple logistic regression models. For this analysis, dichotomized versions of the BDST Total score (0 vs >0) and the CBCL scores for the Total, Externalizing, and Internalizing scales (≤55 vs >55) were used. In each case, the social capital index was related positively to the outcome, with P values of .02, .02, .004, and .22 for the BDST Total, the CBCL Total, Externalizing, and Internalizing scores, respectively.
When overall child functioning, as a binary variable (doing well vs doing less well), and the full six-level social capital index were modeled together in a simple logistic regression model, the OR for each unit of increase in social capital score was 1.35 (95% CI = 1.11, 1.63). The ORs and 95% CIs are shown for each of the four LONGSCAN studies in Fig 2.
Table 4 presents the final logistic model for social capital and child well-being controlled for foster care placement, maternal depression, family income, maternal education, and site differences. In this model the OR for the social capital index is statistically significant (P = .02) and essentially unchanged from the unadjusted estimate. The adjusted OR of 1.29 for social capital indicates that in this sample, the addition of any one social capital indicator increases the odds of doing well by 29%; adding any two increases the odds of doing well by 66%.
Family income (P = .04), but not maternal education, was significantly associated with child well-being in the final model. Each $5000 increment of annual household income was associated with a 12% increase in the odds of doing well. Other factors were also important. Children whose maternal caregiver scored above the clinical cutpoint for depression were 73% more likely to be doing less well than those whose maternal caregivers had depression scores in the normal range (P = .06). Out-of-home placement was also highly associated with doing less well (P = .02).
Because of a concern that the analytic strategy of dichotomizing and pooling variables to create the outcome of doing well and the social capital index could have either masked or exaggerated our results, we also constructed models examining these relationships when the variables were used in their continuous forms. A multivariate linear regression model (not shown) examining the BDST and CBCL scores jointly as continuous outcome variables confirmed that social capital, measured jointly by the indicator variables of the social capital index (in their continuous form whenever possible) was a significant predictor (P = .024) of these four outcomes considered simultaneously, after controlling for the covariates and site indicators listed in Table 4.
Even though the children in this study were recruited into LONGSCAN because of their at-risk status, it is still surprising that so few of them (13%) can be classified as free from developmental or behavioral problems at such a young age. Sameroff5 and his colleagues demonstrated that the accumulation of risk jeopardizes a child's development—particularly when few compensatory factors are present to counteract the negative exposures. In this study, we have investigated the extent to which variables conferring social capital might exert a protective influence. Just as Sameroff's study showed that multiple risk factors must be present before a negative impact in child development can be detected, the results of our study show that the likelihood of doing well increases when multiple protective factors are present.
The amount of social capital available to the families in our sample seems to be limited. In one study that defined social capital as someone who would provide emergency help, 92% of the sample reported the availability of this social resource.14 The indicators used in our social capital index are measures of more regular, on-going social relationships that have the potential for conferring not only help in time of need, but also valuable information, normative expectations and sanctions, and positive role models for children. In our study the most prevalent indicator of this type of social capital, neighborhood support, was found among 59% of the families of children doing well and 48% of the other families. The least prevalent indicator, regular church attendance, was found in 37% of the entire sample. The number of families with multiple indicators of social capital was quite small. Only 16% of all study children reported the presence of at least four of the five indicators measured. Our more rigorous definitions and the use of high-risk samples may have diminished the prevalence of social capital below the levels likely to be found in the general population.
Social capital seems to have a beneficial effect across samples. The aggregation of results from different although related studies may be viewed as a meta-analysis that serves to increase statistical power. The approach of the LONGSCAN consortium is methodologically stronger than conventional meta-analyses because of LONGSCAN's common age points, measures, and study procedures. Designing a series of parallel studies can best be described as a prospective meta-analysis in which the pooling across samples is far less troubled than in other meta-analyses because of the comparability of measures, coding, and definitions.
Site variations do not seem to confound or complicate this analysis. We examined our data by sample and found that social capital has crude estimates of effect that were similar in magnitude and direction across all of the studies. Three of the studies, in Seattle, Baltimore, and San Diego, had social capital effect estimates that were closer to the null with CIs which included 1. However, all of the children from San Diego and Seattle had been reported for child maltreatment and few of these children were doing well. Because many of the children in San Diego were in out-of-home placements, the data gathered on social capital, demographic characteristics, and maternal characteristics may describe a relatively new caretaking environment for many of these children. For foster children, the association between their well-being and their caregiver's social capital will necessarily be confounded by issues related to residential instability and perceptions of impermanent relationships. The Baltimore site's lack of significance may be related to the relatively low level of behavior problems noted by the mothers in this sample. It is unclear why these mothers should systematically report higher levels of functioning compared with the other LONGSCAN samples. The sample in North Carolina represents a group of families with the greatest opportunity for variation in social capital and it is in this sample that social capital seems to have the strongest impact. The importance of social capital is highlighted in the regression model in which the index emerges as an important predictor of child well-being even after taking site differences, foster care placement, family income, and maternal depression into account.
Social capital is a relatively new theoretical construct that has been variably defined in the few studies that have investigated its importance in child development.1 3 4 Measurement of this multidimensional concept can include a number of factors ranging from features of household composition and aspects of family relationships to community support and affiliation. The best constellation of criteria to measure social capital has yet to be determined. Like others who have investigated social capital, we did not plan our study design or instrumentation with measurement of this construct in mind. Rather, we extracted from our dataset the social capital variables that others have investigated or that seemed theoretically pertinent. The indicators that best discriminated between levels of child functioning were the most direct measures of social capital—church affiliation, perception of personal social support, and support within the neighborhood. The other two variables tested, two-parent family and fewer siblings, are crude indicators of social capital because they only suggest that a child will receive a greater allocation of social resources within the family. Because many of the children in this study have troubled families and complicated residential circumstances, these indicators may have had diminished power in this study. As we continue to follow these families, we will attempt to sort out the dimensions of family social capital, as well as take a look at the perceived social support and organizational affiliation of the children themselves. Further study of how individual indicators of social capital exert their influence is also warranted.
The first studies on social capital examined its impact on completion of high school. Our findings suggest that social capital may have an impact on children's well-being as early as the preschool years. In these years it seems to be the parents' social capital that confers benefits on their offspring, just as children benefit from their parents' financial and human capital. Social capital may be most crucial for families who have fewer financial and educational resources. In fact, earlier studies of high-risk neighborhoods and high-risk families have noted that troubled families seem to be more strongly influenced by the neighborhood context than successful families.15 A recent investigation by Korbin and Coulton16 that focuses on children and their neighborhoods concludes that not all poor neighborhoods are alike. The features that characterize the neighborhoods where better functioning families reside are features of social capital: community investment, trust, and organizational affiliation.
Putnam2 has raised concern that social capital is eroding in this country—across social classes and income levels—with detrimental effects on children, families, neighborhoods, and even our democratic political structure. Pediatricians and child advocates may wish to focus attention on the ways in which public policy and service interventions promote or jeopardize the formation of social capital. Our findings suggest that those interested in the healthy development of children, particularly children most at risk for poor developmental outcomes, can intervene to reduce isolation and nurture interpersonal relationships in a variety of ways. Indeed, simultaneous interventions in a number of areas may be necessary to have an effect on child well-being.
The Consortium of Longitudinal Studies on Child Abuse and Neglect has been supported by National Center on Child Abuse and Neglect grants #90CA1433 and #90CA1467 to the University of North Carolina Injury Prevention Research Center, #90CA1481 to the University of Maryland School of Medicine; #90CA1568 to the Juvenile Protective Association of Chicago; and #90CA1566 to San Diego State University. The analysis of social capital and child well-being has been supported by grant UO1HD30945 from the National Institute of Child Health and Human Development.
Significant contributions to the conceptualization and execution of the LONGSCAN initiative have been made by the following investigators: Maureen M. Black, PhD (University of Maryland); Patrick A. Curtis, PhD (Child Welfare League of America, Washington, DC and Juvenile Protective Association, Chicago); Mark D. Everson, PhD (University of North Carolina); Jonathan B. Kotch, MD, MPH (University of North Carolina); Alan Litrownik, PhD (San Diego State University); Mary Wood Schneider, PhD (Juvenile Protective Association, Chicago); and Raymond Starr, PhD (University of Maryland Baltimore County). Professional staff who have been crucial to the successful execution of the study include: Elizabeth Knight, MSW, Ellen Ruina, MBA, Jane Winsor, Hope Bryan, MS, and Shelley Simpson at the University of North Carolina at Chapel Hill; Donna Harrington, PhD, at the University of Maryland; Jean Remmer at San Diego State University; and Edie Nelson, MSW, at the Washington State Department of Social Services. The manuscript has benefited from thoughtful comments from Drs Carol Runyan and Keith Wailoo at the University of North Carolina and the members of the National Institute of Child Health and Human Development Family and Child Well-being Network.
- Received November 11, 1996.
- Accepted May 22, 1997.
Reprint requests to (D.K.R.) Department of Social Medicine, CB# 7240, University of North Carolina School of Medicine, Chapel Hill, NC 27599–7240.
This article comes from the Consortium for Longitudinal Studies of Child Abuse and Neglect (LONGSCAN) and the NICHD Family and Child Well-Being Network.
Early versions of this article were presented at the Population Association of America Annual Meeting in San Francisco, California, April 5, 1995, and the Third International Conference on Injury Prevention and Control in Melbourne, Australia, February 21, 1996.
- Smith MH,
- Beaulieu LJ,
- Israel GD
- Sameroff AJ,
- Seifer R,
- Baroces R,
- Zax M,
- Greenspan S
- ↵Runyan DK, Curtis P, Hunter WM, et al. LONGSCAN: A consortium for longitudinal studies of maltreatment and the lifecourse of children. Aggression and Violent Behavior: A Review Journal. In press
- ↵Newborg J, Stock JR, Wnek L, Guidubaldi J, Svinicki J. Battelle Developmental Inventory Screening Test. Allen, TX: DLM-Teaching Resources; 1984
- ↵Achenbach TM. Manual for the Child Behavior Checklist/4–18 and 1991 Profile. Burlington, VT: University of Vermont Department of Psychiatry; 1991
- ↵SAS Institute Inc. SAS/STAT Software: Changes and Enhancements. Release 6.10. Cary, NC: SAS Institute Inc; 1994
- ↵Korbin J, Coulton C. Neighborhood Impact on Child Abuse and Neglect: Final Report on Grant #90CA-1494. Washington, DC: National Center on Child Abuse and Neglect; 1995
- Copyright © 1998 American Academy of Pediatrics