The final step of the analysis was to determine whether these associations persisted when we carried out multivariate regressions. This statistical procedure makes it possible to examine the association between a pair of variables (such as mother’s level of education and the amount of time children spend hanging out) while simultaneously holding constant (i.e., controlling for) other variables, such as age, race, or income. The goal here is to determine whether the association between the variables of interest exists among individuals who are also comparable to one another in other important respects. So, if we are concerned with the relationship between the amount of education mothers have and the amount of time their children spend hanging out, we would want to control for various factors that might also be related to hanging out—things such as age, gender, race, family structure (single-parent or two-parent), mother’s work status, and so forth. Essentially, multivariate regression analysis enables us to isolate the predictive power of mother’s education.
In the first of these analyses, we found that two of our proxy measures of social class—mother’s education and family income—remain powerful in predicting children’s participation in organized activities, despite the inclusion of numerous control variables. The accompanying figures indicate the magnitude of the relationships we found. For illustration, we present the predictions our statistical model makes for a hypothetical Black boy (Figure 1) and white girl (Figure 2). (To facilitate comparison, these hypothetical children are assumed to be identical in all other respects captured by the model—i.e., in terms of family structure, age, wealth, etc.) The patterns throughout the figures are quite similar. Indeed, we found that, with the control variables included, there is no discernable difference between Black and white children in participation in organized activities. These results confirm some of the central findings from Unequal Childhoods: participation in organized activities appears to be closely linked to social class, but not to race. Furthermore, while the effects revealed in the graphs are fairly substantial in magnitude, it is worth bearing in mind that income and education frequently “overlap” with one another—that is, families like the ones described as “middle-class” in Unequal Childhoods tend to have both high incomes and high levels of educational attainment. Thus the graphs should be viewed as conservative estimates of social class differences in organized activity participation.5
The multivariate analysis of hanging out also shows a pattern consistent with the findings of Unequal Childhoods. Children’s participation in non-organized leisure is closely associated with their mother’s level of education. Here, as in the first edition, the relationship is negative, meaning that children whose mothers have higher amounts of education tend to spend less time hanging out than children whose mothers have lower levels of education. Thus, it would appear that in middle-class families, organized activities substitute for hanging out. And, again, we found no evidence of a difference between Black and white children.
Note: Based on coefficients from a tobit regression. These computations assume a reference child who is Black and male; whose family wealth is in the second quartile; whose father works and whose mother does not; who has the sample mean values of age and number of siblings; and whose time diaries were collected on Wednesday and Saturday. For the education charts, income is set to the second quartile; for the income calculations, education is set to high school. Source: Based on author’s analysis of PSID-CDS data.
Figure 1. The effects of maternal education and family income on children’s participation in organized activities, for a reference child who is Black and male.
Note: Based on coefficients from a tobit regression. These computations assume a reference child who is white and female; whose family wealth is in the second quartile; whose father works and whose mother does not; who has the sample mean values of age and number of siblings; and whose time diaries were collected on Wednesday and Saturday. For the education charts, income is set to the second quartile; for the income calculations, education is set to high school. Source: Based on author’s analysis of PSID-CDS data.
Figure 2. The effects of maternal education and family income on children’s participation in organized activities, for a reference child who is white and female.
In addition to the core argument concerning class differences in time use, language use, and parents’ intervention in institutions, a number of other findings are threaded through Unequal Childhoods. One is that the middle-class children have less contact with their relatives than do working-class and poor children. The ethnographic research suggests that this finding is true for both whites and Blacks. In the case of time spent with extended kin, however, the multivariate results proved somewhat surprising.6 To be sure, the level of a mother’s education is a powerful predictor: consistent with the ethnographic data, we found that as maternal education increases, contact with relatives decreases. Thus, in a representative sample of U.S. children, middle-class children spend less time with their relatives than do working-class children. However, race is also significant here: Black children are considerably more likely than their white counterparts to have contact with extended kin. Indeed, the difference between Black and white children is roughly comparable in magnitude to the difference between children whose mothers dropped out of high school, on the one hand, and those whose mothers completed a bachelor’s degree, on the other. So, although the social class pattern is sustained, the PSID-CDS data reveal a race-related difference in the amount of time white and Black children spend with their relatives, a difference that did not show up in the ethnography.
Overall, the PSID-CDS analyses indicate that at a national level, differences between families exist that generally accord with the results of Unequal Childhoods. Children of highly educated parents and of high-income parents exhibit substantially greater involvement in organized activities than their peers whose parents are less well-educated and have lower income. Conversely, the children of less educated parents spend far more time hanging out than their counterparts with highly educated parents. These results are broadly consistent with the premise of an American middle class that tends to engage in concerted cultivation and of a working class and a poor group that tend to engage in the accomplishment of natural growth. The quantitative research also showed an interesting departure from the ethnographic work: the class pattern of time spent with relatives holds, but a significant race difference is also apparent.