The Family Life Project, Phase I, United States, September 2003-January 2008 (ICPSR 34602)

Version Date: Feb 14, 2022 View help for published

Principal Investigator(s): View help for Principal Investigator(s)
Lynne Vernon-Feagans, University of North Carolina at Chapel Hill; Ann C. (Nan) Crouter, Pennsylvania State University; Martha J. Cox, University of North Carolina at Chapel Hill

Version V5 ()

  • V5 [2022-02-14]
  • V4 [2018-06-27] unpublished
  • V1 [2015-02-26] unpublished
Slide tabs to view more

The purpose of this project is to study the early development of a group of children who are at risk regarding later successful adjustment and for whom we have little information: children living in rural, largely poor communities. A birth cohort of 800 children in three rural counties of North Carolina and 600 children in three rural counties of Pennsylvania were studied over the first 3 years of life. A multidisciplinary team investigated multiple levels of influences affecting the early development of these children.

The research emphases of the component research projects include the following:

  • Project I: Temperament: Emphasizes the development of child-related factors and how they predict preschool social-emotional and cognitive competence
  • Project II: School Readiness: Emphasizes the pathways to and precursors of school readiness
  • Project III: Family Process: Emphasizes how family processes mediate or moderate the effects of rural poverty on children
  • Project IV: Work and Family: Emphasizes the impact of parents' occupational conditions on parenting, and, in turn, children's social, cognitive, emotional and linguistic development
  • Project V: Ethnography: Emphasizes two components:
  1. an in-depth contextual appraisal of community characteristics
  2. a family ethnography with 72 families developmentally ahead of the cohort above to provide input to design and measurement

This project used a cumulative risk model to examine the relation between social risk and children's executive functioning, language development and behavioral competence at 36 months. Using both the Family Process Model of development and the Family Investment Model of development, observed parenting was examined over time in relation to child functioning at 36 months. Different aspects of observed parenting were examined as mediators/moderators of risk in predicting child outcomes. Results suggested that cumulative risk was important in predicting all three major domains of child outcomes and that positive and negative parenting and maternal language complexity were mediators of these relations. Maternal positive parenting was found to be a buffer for the most risky families in predicting behavioral competence. In a final model using both family process and investment measures, there was evidence of mediation but with little evidence of the specificity of parenting for particular outcomes. Results demonstrate the importance of cumulative risk and parenting in understanding child competence in rural poverty and the implications for possible intervention strategies that might be effective in maximizing the early development of these children.

Vernon-Feagans, Lynne, Crouter, Ann C. (Nan), and Cox, Martha J. The Family Life Project, Phase I, United States, September 2003-January 2008. Inter-university Consortium for Political and Social Research [distributor], 2022-02-14.

Export Citation:

  • RIS (generic format for RefWorks, EndNote, etc.)
  • EndNote
United States Department of Health and Human Services (PHS 398)


Access to parts of this study requires a signed User Agreement. To obtain the restricted files, researchers must agree to the terms and conditions of the Restricted Data Use Agreement, found via ICPSR's online Restricted Data Access Request System, by clicking the "apply online for access to the data" link above.

Inter-university Consortium for Political and Social Research

2003-09-15 -- 2008-01-31
2003-09-15 -- 2008-01-31 (Multiple time points based on child's age)
  1. Restricted-use files for tracking phone call data, child care interview data, and interview grid data are now available via a signed User Agreement and online application.

    In addition, public codebooks listing variables and variable descriptions for primary respondent interview and assessment data, secondary respondent interview and assessment data, GIS block group and tract-level contextual data, and GIS coding data. Frequencies in these codebooks are suppressed. Release of these files is forthcoming.

  2. Additional information about this project can be found on the Family Life Project Web site.


About 20 percent of children in the United States have been reported to live in rural communities, with child poverty rates higher and geographic isolation from resources greater than in urban communities. There have been surprisingly few studies of children living in rural communities, especially poor rural communities. The Family Life Project helped fill this gap by using an epidemiological design to recruit and study a representative sample of every baby born to a mother who resided in one of six poor rural counties over a one year period, oversampling for mothers living in poverty and African American mothers. 1,292 children were followed from birth to 36 months of age.

All families living in the six target counties who have given birth to a child during the recruitment period (1 year) were eligible to participate in the project. Mothers were screened in the hospital and selected according to their income status in Pennsylvania and race/income status in North Carolina.

Data collection consisted of a series of home visits, child care visits, and phone calls. Families were visited when the infant was 2, 6, 15, 24, and 36 months of age. The 2- and 15-month assessments consisted of a single home visit that involved the primary caregiver and the target child. If available and willing, the secondary caregiver was also included at the 15-month assessment. The 6-, 24-, and 36-month assessments consisted of two separate home visits that involved the primary (and, when possible, the secondary) caregiver, as well as the target child. A majority of home visitors resided in the counties in which the families lived. In North Carolina, African American home visitors were selected to make sure all home visits to African American families included a home visitor of the same race. Two home visitors visited the families at each time point in order to collect the data in a timely fashion and to ensure that all tasks were completed successfully. At the 6-, 15-, 24-, and 36-month assessments, visits were made to the children's most-used child care setting if the child was in any form of child care for a total of 10 hours a week or more. Phone calls were made to each family approximately every four months between designated assessments, in order to gather updated information about family moves.

Home visits lasted approximately 2-3 hours and consisted of a variety of interviews, questionnaires, videotaped interactions between the primary caregiver and the child, the secondary caregiver and the child, and direct child assessments. Questionnaire, interview, and direct assessment data were collected on laptops. To ascertain reading levels, all primary and secondary caregivers completed the KFAST literacy screener (Kaufman and Kaufman, 1994) during their first visit. Participants who read at an 8th grade reading level or above were given the opportunity to complete questionnaires on their own (86 percent sample), whereas those who read below an 8th grade reading level had questionnaires read to them by one of the interviewers. For confidential information, informants were given a key pad so that the interviewer would not know the answer to the question.

Child care visits were conducted after contacting families to obtain the name and address of the child care person and to get written permission to visit the child care setting. We visited 72 percent of the child care settings over the course of the first three years. The setting could be in the child's home or another home or formal child care. The caregiver could be a relative, babysitter, biological non-residential father, residential grandmother who was not a primary or secondary caregiver, and other related or non-related adults. Child care visits were completed by interviewing the caregiver about the setting and her/his background characteristics and through observations of the setting over about a one and a half hour time period.

Sampling was done to obtain a representative sample (i.e., a birth cohort) of children born in one of six targeted counties between September 2002 and September 2003. Low-income families and African American families (North Carolina only) were over-sampled. A broad definition of rural was used, which included counties with small to medium sized towns with little access to large metropolitan areas and the resources those areas might provide. There are four areas of the United States that have been previously identified as having high rates of rural poverty (Dill, 1999) and this study sampled from two of these areas, Appalachia and the African American South.

County inclusion criteria was: counties with no town with a population of greater than 50,000 and counties where nearly half the elementary age children were eligible for free and/or reduced lunch (children would be within 180 percent of the national poverty line). Counties that were adjacent to large metropolitan areas were excluded, so that target counties could not be considered suburban. Counties that were remote and completely rural were also excluded. Using this definition, 28 counties in Pennsylvania and 40 in North Carolina met our definition of "rural" and poor, with Beale codes of 3, 4, and 6. Within these counties, geographic clusters were identified to reduce data collection and recruitment costs. Demographic factors (poverty and African American race) were also considered. Six counties meeting the above criteria were chosen.

Next, families were sampled from hospitals. In the Pennsylvania counties, there were seven hospitals that delivered babies. Given prior hospital administrative data regarding the number of births expected to occur, four of the seven hospitals were randomly selected to participate. In the North Carolina counties, there were only three hospitals that delivered babies so further sampling was not needed. There were additional considerations for sampling of the families including representative sampling for high-risk births, and attention towards the number of days a week and which days a week families were sampled. Three criteria were used to exclude families from participation: non-English speaking families, families intending on moving within the next three years (such as, military enlistment), and families where the state had terminated parental rights (such as, women in prison) were all excluded. Families were classified into low income status based on self-reports including social service use, highest education achieved, and household income, which served as inclusion criteria. In North Carolina, race was also considered as inclusion criteria due to the oversampling of African American babies, whose race was designated by mothers.

Longitudinal: Panel, Longitudinal, Longitudinal: Panel: Continuous, Longitudinal: Cohort / Event-based

Rural children living in poverty in eastern North Carolina and central Pennsylvania.

Individual, Household

Information from home visits, childcare visits, and phone calls was collected on the following domains of interest:

  • Attention
  • Adult Literacy and Cognition
  • Adult Relationships
  • Child Behavioral Measures
  • Child Cognitive Measures
  • Child Temperament
  • Child Care Information
  • Community and Neighborhood Environment
  • Demographics
  • Emergent Literacy
  • Executive Functioning
  • Genotyping Data
  • Home Environment
  • Income and Assistance
  • Language Development
  • Parent Mental Health
  • Parenting
  • Physical Health
  • Rurality
  • Stress
  • Support
  • Work and Job Characteristics

The Family Life Project enrolled 1200 families total. In Pennsylvania, 320 of these families were low income, while 160 were not low income, for a total of 480 families. In North Carolina, 540 families were low income (380 of these were African American), while 180 were not low income (100 of these were African American) for a total of 720 families (480 African American families). Overall, there were 860 families that were low income and 340 families that were not low income.

Given successful retention efforts, a low attrition rate of 6.19 percent was maintained through 36 months of age of the target child.

In most cases, each data file represents individual responses on a scale. Some data files contain responses from individuals on the same scale at multiple time points. In some instances, there are separate data files for the same scale at different time points.



2022-02-14 This study was updated to include the Wechsler Preschool and Primary Scale of Intelligence III (WPPSI) Restricted-Use Data set, which is a restricted version of the Wechsler Preschool and Primary Scale of Intelligence III (WPPSI) Data set, and includes the variable 'S_ID'.

2018-06-27 This study was updated to include public ddi files for datasets 2, 5, 6, 10, 11, 15, 16, 21, 22, 51, and 52.

2018-02-15 The citation of this study may have changed due to the new version control system that has been implemented. The previous citation was:
  • Vernon-Feagans, Lynne, Ann C. (Nan) Crouter, and Martha J. Cox. The Family Life Project, Phase I, United States, September 2003-January 2008. ICPSR34602-v5. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2022-02-14.

2015-02-26 Restricted-use data files for household tracking information, child care interviews, and summary demographic information were released. Public codebooks listing variables and variable descriptions for files containing geographic information and detailed household-level information were also released.

2014-10-13 Update to drop a variable and make a labelling correction to the weights file.

2014-09-30 ICPSR data undergo a confidentiality review and are altered when necessary to limit the risk of disclosure. ICPSR also routinely creates ready-to-go data files along with setups in the major statistical software formats as well as standard codebooks to accompany the data. In addition to these procedures, ICPSR performed the following processing steps for this data collection:

  • Performed consistency checks.
  • Created variable labels and/or value labels.
  • Standardized missing values.
  • Created online analysis version with question text.
  • Checked for undocumented or out-of-range codes.

The Weights data file (part 0095) contains all weight variables created for use with analysis. This file can be merged by s_id, and identification number by family.

SUDAAN was used to create the weight variables. "With replacement" sampling is used based on the project's interest in making inferences about low income/African American families in Appalachia and the African American South, as opposed to this particular set of families. The sampling design is a stratified sample with unequal weighting (state, county, and hospital were not sampled).

Weights were created to take into account non-response and oversampling to achieve target numbers. Data were weighted separately for each hospital and stratum to create the probwght variable. An 18 level variable was created corresponding to hospital, race, and income called the strat variable. There were 7 levels for each hospital that was sampled, 2 levels for race (African American vs not African American), and 2 levels for income (low income or not low income). Stratification variables for state, ethnicity, and low income level are also included.