«Demographic Research a free, expedited, online journal of peer-reviewed research and commentary in the population sciences published by the Max ...»
6.2 Effects of single parenthood on educational aspiration Table 3 reports estimates obtained from four logistic regression models of predicting whether the respondent aspire to four-year university education among middle school and high school seniors combined.7 According to Model 1, net of parental education, gender, and number of siblings, living with a divorced single father decreases the log odds of aspiring to four-year university education by 1.042. Stated differently, the odds of aspiring to four-year university education among students with a divorced single father are 0.35 times the odds among students with two parents (e−1.042 = 0.35), holding gender, number of siblings, and parental education constant. The odds among students with a divorced single mother are a half of the odds among students with two parents (e−0.685 = 0.50).
In contrast to the comparison between two-parent and divorced single-parent families, differences in educational aspiration between students with two parents and those with a widowed single father and with a widowed single mother are not statistically signiﬁcant.
7 I ﬁrst estimated the models separately for middle school seniors and high school seniors.
The results showed that relationships between each type of single-parent family and educational aspiration were very similar between middle school seniors and high school seniors. I also could run the models only among high-school seniors by including the type of high school (academic or vocation schools) as an independent variable predicting the likelihood of aspiring to four-year university education. Studies showed greater likelihood to attend vocational schools among students from disadvantaged background (Phang & Kim, 2003). Considering that single parenthood is associated with poorer economic conditions and attending vocational schools constrain prospects for university education, the type of high school may mediate the effect of growing up in a single-parent family on educational outcomes. However, because I am interested in the total effect of single parenthood and the extent to which the total effect is explained by household income and parenting practices, I did not include the school type in the models.
Earlier, descriptive statistics in Table 2 showed that differences between students with two parents and those with a widowed single parent were substantial before gender, number of siblings, and parental education were controlled.8 The additional analyses (not shown) revealed that among the three control variable, parental education was most relevant in explaining the difference between children from two-parent families and children from a widowed single parent.
In Models 2 through 4, I examine the extent to which household income and parentchild interaction account for the differences by family type revealed in Model 1, by adding to Model 1 household income (Model 2), four indicators of parent-child communication (Model 3), and both (Model 4). In order to facilitate interpretation of results in Table 3, I transformed the coefﬁcients to the odds ratios (Figure 2). The odds ratio of 1 indicates no difference in the likelihood of aspirating to four-year university education between students with two parents and those with a single parent compared. The odds ratio less than 1 indicates that students from single-parent families are less likely to aspire to fouryear university education than students from two-parent families. As the odds ration becomes closer to 1, the difference in educational aspiration between students from twoparent families and those from single-parent families becomes narrower.
The ﬁrst set of bars for the difference between students with a divorced single father and students with two parents shows that by controlling for household income, the odds ratio increases from 0.35 to 0.44, indicating the decreasing gap between students from the two types of families. Controlling for parent-child interaction in addition to gender, number of siblings, and parental education increases the odds ratio as much as does controlling for household income. Finally, the odds ratio increases from 0.35 in Model 1 to
0.53 in Model 4 when both household income and parent-child interaction are taken into account. Note, however, that the lower level of educational aspiration among students from divorced single-father families remains still signiﬁcant (see Table 3).
The next set of bars shows the change in the odds ratio for the comparison between students with a widowed single father and those with two parents. After either household income or parent-child interaction is controlled, the odds ratio becomes greater than 1 indicating that students with a widowed single father are more likely to aspire to fouryear university education. However, the small number of students with a widowed single father (N= 43) renders the difference statistically non-signiﬁcant, requiring a caution in interpreting the result. An interesting observation, though, is that the extent to which parent-child communication mediates the effect of single fatherhood due to divorce seems 8 Notethat without the three variables controlled the difference between students with two parents and those with a widowed single mother was statistically signiﬁcant. Although the difference between students with two parents and those with a widowed single father was large enough (the coefﬁcient was -0.618), it was not statistically signiﬁcance because of the large standard error caused by the small number of students with a widowed single
greater than does household income as indicated by the larger degree of change in the odds ratio.
The change in the odds ratio for the comparison between students with a divorced single mother and students with two parents is presented in the third set of bars. Controlling for household income increases the odds ratio from 0.50 in Model 1 to 0.70 to Model 2.
It indicates a signiﬁcant role of economic insecurity in explaining the lower level of educational aspiration among students from divorced single-mother families. Contrastingly, the extent to which parent-child interaction accounts for the lower level of educational aspiration among students with a divorced single mother is negligible. The less likelihood of aspiring to four-year university education among students from divorced single-mother families remains signiﬁcant at the 0.10 level even after both household income and parentchild interaction are taken into account (see Table 3).
Finally, in the comparison between students with a widowed single mother and students with two parents, the odds ratio becomes greater than 1 after household income is included in addition to gender, number of siblings, and parental education. However, the different is not statistically signiﬁcant (Table 3). Similar to the result for student with a
divorced single mother, parent-child interaction does not seem to play a signiﬁcant role in accounting for the effect of single motherhood due to the death of a father (the odds ratio hardly changes).
6.3 Effects of single parenthood on school disengagement Table 4 presents the results for student disengagement as measured by school absence and punishment. Model 1, which includes only the three control variables, shows that students from divorced single-father families and from divorced single-mother families are more likely to be disengaged compared to students from two-parent families. Although the coefﬁcient of 0.570 associated with single fatherhood due to the death of a mother seems to be substantial in the size, it is not statistically signiﬁcant reﬂecting the small sample size.
Students living with a widowed mother do not show signiﬁcant difference in disengagement as compared to those living with two parents.
In Figure 3, the coefﬁcients in Table 4 are presented as the odds ratio across models.
Even after both household income and parent-child interaction are taken into account, in
addition to gender, the number of siblings, and parental education, the higher likelihood of disengagement for students with a divorced father or a divorced mother as compared to those with two parents remains signiﬁcant. Controlling for those variables does not change non-signiﬁcant differences between students with two-parents and those with a widowed parent.
In terms of the relative roles that household income and parent-child interaction play in accounting for the differences in student disengagement, notable is the relative importance of parent-child interaction over household income among students with a single father. For instance, the comparison between students with a divorced single father and students with two parents shows that the odds ratio decreases from 2.52 in Model 1 to
2.35 in Model 2 with household income additionally controlled, while the odds ratio decreases from 2.52 in Model 1 to 2.09 in Model 3 with the four indicators of parent-child interactions controlled. The same pattern is found among students with a widowed single father, although the effects of single fatherhood due to the death of a mother are not statistically signiﬁcant. Contrastingly, parent-child interaction accounts for little of the difference between students with a widowed single mother and students with two parents.
The relative explanatory power between household income and parent-child interaction among students with a divorced mother is similar. It is worth mentioning that the extent to which the two mediating variables, particularly household income, explain the effects of single parenthood on student disengagement is substantially smaller than the effects on educational aspiration. Indeed, adding household income to Model 1 to predict student disengagement does not signiﬁcantly improve the model ﬁt at the 0.05 level.9
6.4 A supplementary analysis The existing literature of single parenthood and its consequences for children’s well-being has not systematically examined the potential impact of the extended family system for moderating the negative consequences of growing up with a single parent. The ignorance may be partially attributable to the relatively small size of single-parent families with grandparent present in the United States and some Western societies. Considering that substantial numbers of children in single-parent households live with grandparents in Korea, the question of how the extended family system, especially co-residence with grandparents, mediates the effects of single parenthood on children’s education can be of interest. The hypothesis of particular interest is that having at least a grandparent present in households should be associated with better educational outcomes of children and the effect of grandparents should be strong especially for single-parent families.
9 Thelikelihood ratio statistic (G2 ) for model comparison is 4.9 with 2 degrees of freedom (6280.6 in Model 1 in Model 2). G2 is asymptotically distributed as chi-square with the degrees of freedom. The P value associated with G2 = 4.9 (with 2 df) is between 0.05 and 0.10.
However, the fairly small size of sample used in the current study prevented a serious examination of the issue. Sample sizes for single-parent families were considerably small even before being separated by the presence of grandparent (refer to unweighted Ns in the bottom of Table2). There were only 10 students who lived with a widowed single father and at least a grandparent. Similarly, the sample size of students living with a widowed single mother and at least a grandparent was only 14. Therefore, I ﬁrst looked at the simple bivariate relationships between co-residence with grandparents and educational outcomes. For most types of family, co-residence with grandparent showed either no signiﬁcant relationships or even slightly negative relationships with educational outcomes.
Despite the sample sizes, furthermore, I conducted multivariate analyses that included interaction terms between the presence of grandparents and family structure. The results (not shown here) revealed that for children with two parents the effect of co-resident grandparent was negligible, and none of the interaction effects between co-resident parents and family types were statistically signiﬁcant because of large standard errors caused by the small sample sizes. More importantly, similar to the bivariate results, the direction of interaction effects mostly indicated that children of single parents who had at least a grandparent in households were actually worse off than children of single parents who did not have a grandparent present.10 Although the negligible or negative effect of living with a grandparent on children’s education is somewhat unexpected, the pattern has already been reported in several studies with cross-sectional data. Examining the effect of a co-resident grandparent on 15-yearold students’ test scores in 30 countries, Marks (2007) found the negative association between having a co-resident grandparent and test scores in most countries. Another cross-national study of 4th and 8th graders’ math scores by Moyi, Pong, and Frick (2004) also showed that the negative association was predominant among 25 countries and moreover, the pattern was consistent across different family types.