Food insecurity in early childhood is especially concerning because the first 5 years of a child’s life are a critical period for brain development, and deprivation during these early years has long-lasting consequences. Unfortunately, 20 percent of U.S. households with children experience food insecurity; nearly half had children who experienced food insecurity directly. Prior research suggests that older children who experience food insecurity may face an increased risk of cognitive and social deficits, which present obstacles to later school and life success. However, it is unknown how the intensity and duration of food insecurity experienced in the earliest years (birth to age 5) might relate to children’s kindergarten cognitive and social outcomes, which are themselves predictors of later achievement and earnings. Separately, it is unknown how parents of young children might combine multiple food-assistance programs that have the potential to reduce early childhood food insecurity and, subsequently, achievement disparities.
Understanding patterns and predictors of multiple food-assistance-program use could illuminate promising pathways for reducing food insecurity and its attendant achievement disparities. This study addressed these two important but unanswered questions, in two separate studies, asking (1) what is the association between food insecurity in early childhood and kindergarten school-readiness outcomes?, and (2) what are the patterns and predictors of multiple food assistance program use among families with young children across their child’s earliest years?
These questions are addressed using data from the Early Childhood Longitudinal Study–Birth Cohort (ECLS-B). The ECLS-B is a nationally representative study of children born in 2001; parents were interviewed about their food-assistance-program use and food insecurity (using all 18 items of the USDA’s Core Food Security Module) and children were directly assessed in 2001 when children were 9 months old, in 2003 when they were 2 years old, in 2005 to 2006 when they were in preschool, and in 2006 to 2007 when they were in kindergarten. At the kindergarten wave, in addition to direct cognitive assessments, kindergarten teachers reported on children’s social-emotional functioning. The key variables for this study constituted use of the Special Supplemental Food and Nutrition Program for Women, Infants, and Children (WIC), the Supplemental Nutrition Assistance Program (SNAP), household food insecurity, and children’s directly assessed reading and math scores and their kindergarten teacher-reported levels of hyperactivity, conduct problems, and approaches to learning. Across both studies, the ECLS-B sample was restricted to families with income at or below 185 percent of the Federal poverty level.
STUDY 1: To assess associations between food insecurity intensity and duration across early childhood and children’s kindergarten cognitive and social-emotional outcomes, ordinary-least-squares regression models were used. Separate models were estimated for each outcome, and all models included a comprehensive set of family- and child-level covariates (maternal education, employment, average income [from Waves 1–3], household receipt of public benefits, race, immigrant status, English proficiency, maternal depression, mother’s age at child’s birth, number of people in the household, number of children under age 6 in the household, number of children in the household over age 7, urbanicity, childcare use, child gender, child age, kindergarten entry year, and child special needs status). For models predicting kindergarten outcomes from food insecurity at child age 9 months, no lagged outcome was used because 9 months represented the first data collection wave. However, for models estimating kindergarten outcomes from food insecurity at child age 2 years and preschool, separately, the lagged outcome from the prior wave was used.
STUDY 2: To identify patterns of multiple food assistance program use across the early childhood years, latent class analysis (LCA) was used. LCA is an empirical approach that can be used to illuminate distinct population subgroups according to the most common combinations of variables. In the case of this study, LCA was used to identify profiles of WIC and/or SNAP use across each of the ECLS-B’s four waves (9 months, 2 years, preschool, and kindergarten). This yielded eight possible use points, that is, two programs at each of four data collection waves. After identifying profiles of multiple food-assistance-program use, multinomial logit regression was then used to predict membership in each profile group. The same covariates used in Study 1 as controls were used as predictors of group membership in Study 2.
STUDY 1 FINDINGS: At 9 months, very low food security was associated with increased hyperactivity and conduct problems, and reduced reading scores; low food security was related to decreased math scores. At 2 years, very low food security was again associated with increased hyperactivity and conduct problems, and with decreased approaches to learning and math skills; low food security was also linked to decreased reading and math performance. At preschool, very low food security was significantly linked with decreased approaches to learning only, while low food security was linked with reduced reading scores. As the number of episodes of food insecurity increased across early childhood, so did the size of the negative effect on kindergarten outcomes.
STUDY 2 FINDINGS: LCA revealed six distinct food-assistance-program use groups, five of which were sufficiently populated. These five groups included: the never users, or families who never used WIC or SNAP (or only used one program, once, out of eight possible use points); the always users, or families who used WIC and/or SNAP at each of the eight possible use points; those who transitioned from WIC and SNAP (using WIC and/or SNAP for the first two waves) to just SNAP; those who transitioned from WIC (using WIC for the first two waves) to no food assistance; and the heavy WIC users, or those who used WIC at three or more of the four waves. There was a sixth class that constituted roughly 5 percent of the sample for whom there were no discernable patterns and which was excluded from analysis. Next, multinomial logit models were used to predict group membership from covariates. Not surprisingly, the always users were the most disadvantaged (for example, most likely to be non-white, to have lower incomes, and to have mothers who reported emotional distress), followed by the heavy WIC users. Those who transitioned from WIC to SNAP were slightly less disadvantaged, whereas those who transitioned from WIC to no assistance looked most like the most advantaged group, that is, the never users.