fraction of individuals who are poor, the fraction of adults with a college degree, and the fraction of adult men without jobs. Such data are available for census tracts (geographic areas encompassing 4,000 to 6,000 individuals, with boundaries drawn to approximate neighborhood areas) as well as larger geographically defined areas.

One striking result in broad-based studies of neighborhood effects on young children is that there are many more differences in families and children within neighborhoods than between them. As a result, in one study, neighborhood factors such as poverty, male joblessness, and ethnic diversity were found to account for only a small share of the differences across 5- to 6-year-old children in problem behaviors and academic achievement (Klebanov et al., 1997, Table 4.10). The presence or absence of affluent, high-income neighbors, rather than of poor neighbors, related more strongly to child and adolescent outcomes. This may not be a direct effect of income per se; it may derive from the differing social and interpersonal resources that are available in higher-income neighborhoods, as emphasized in social organization theories of neighborhood influence, as well as their greater support for sustaining academic achievement and other positive efforts (Darling and Steinberg, 1993). Yet neighborhood conditions added at most 2 percent to the explained variation in young children's outcomes over and above family conditions.

Neighborhood factors also do not account for much of the variation in parental mental health and family management practices. Using data from the Infant Health and Development Program, Klebanov et al. (1994) found that at most 4 percent of the variation in the depression, social support, and behavioral coping of mothers of preschoolers could be accounted for by neighborhood conditions. With data from a diverse set of Philadelphia neighborhoods, Furstenberg et al. (1999: Table 7.1) found similar results for their measures of the psychological resources of adolescents' care-givers—fully 90 percent of the variance in family management practices was found within rather than between neighborhoods.

These results indicate that even if we could somehow equalize neighborhood conditions, it would have little impact on the dispersion of family mental health and management practices or on individual differences in children's behavior problems and achievement. However, we caution against drawing more practical policy conclusions from these patterns of explained variance (Cain and Watts, 1972; Duncan and Raudenbush, in press; Rosenthal and Rubin, 1982). The cost-effectiveness of a neighborhood intervention depends on effect sizes relative to cost, and socially profitable intervention policies are quite possible in the context of a small amount of explained variation.

The neighborhood study of Sampson, Raudenbush, and Earls (1997) is noteworthy for its focus on the “collective efficacy” of neighborhoods. This



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement