Well-Being of Children in Working Poor and Other Families: 1997 and 2004. Multivariate findings

09/01/2008

Findings for the multivariate analyses of the dichotomous well-being measures are displayed in Table 4 for the logistic models.  Models with no controls are displayed in the first column.  Models with basic demographic controls are displayed in the second column.  Models with basic demographic controls and the work pattern variables are displayed in the third column.  We display the odds ratio for income below the poverty line and not making a substantial work effort with working poor as the omitted category.  Findings for the ordinary least squares analyses of the well-being measures are displayed in Table 5.  We display the OLS coefficient for the income below the poverty line and not making a substantial work effort with working poor as the omitted category.

Logistic models.  As expected, as shown in Table 4, for all outcome measures in which there was a significant difference in the 2004 bivariate analyses between children in working poor families and children in non-working poor families, the odds ratios in the no-control models were statistically significant and in the expected direction.  For example, children in non-working poor families were estimated as only 58 percent as likely as children in working poor families to be enrolled in gifted and talented programs.  The odds ratio was statistically significant at the .01 level.

After controlling for race/ethnicity, family structure, parental education, and parental age, children in non-working poor families were estimated as 64 percent as likely as children in working poor families to be enrolled in gifted and talented programs  a smaller difference, but still statistically significant at the .01 level.  Thus, only a small part of the difference in participation in gifted and talented programs is due to demographic and social differences between children in working poor families and children in non-working poor families.  In addition to enrollment in gifted and talented programs, the following measures were still statistically significant at the .05 level or better after controlling for demographic and social differences:

  • Fathers educational aspirations;
  • Enrollment in private school; and
  • Enrollment in school with religious affiliation.

When controls for full-time/part-time and full-year/part-year work patterns were added, children in non-working poor families were estimated as 75 percent as likely as children in working poor families to be enrolled in gifted and talented programs  a still smaller difference, but still statistically significant at the .05 level.  Thus, an additional part of the difference in participation in gifted and talented programs was due to demographic and work-pattern differences between children in working poor families and children in poor families not making a substantial effort, but a significant difference still remained even after applying all these controls.

The following measures were still statistically significant at the .05 level after controlling for demographic and social differences and differences in work patterns:

  • Fathers education aspirations; and
  • Enrollment in school with religious affiliation.

However, the pattern was somewhat different for fathers educational aspirations.  As expected, among children living in families with their father present, the fathers having educational aspirations for his child to be more education or training after college were only 72 percent as likely for those in non-working poor families than for those in working poor families.  And, as expected, this likelihood increased to 76 percent after controlling for demographic variables.  However, after adding work pattern controls, the difference widened, and children in non-working poor families were only 64 percent as likely as children in working poor families to have a father with educational aspirations for his child to have more education or training after college.

Ordinary least squares models.  Results for the ordinary least squares models followed a similar pattern to the results for the logistic models.  For all outcome measures in which there was a significant difference in the bivariate analyses between children in working poor families and children in non-working poor families, the regression coefficients in the no-control models were statistically significant and in the expected direction. 

For example, in the no-control model predicting parental aggravation, children in non-working poor families were more likely than children in working poor families to have an aggravated parent.  More specifically, the coefficient of income below the poverty line and not making a substantial work effort is -0.321, which was significant at the .01 level.[xi] After controlling for basic demographic and social variables, this coefficient decreased to -0.261 but remained statistically significant at the .01 level.

In addition to parental aggravation, other measures remaining statistically significant at the .05 level or better after demographic and social controls include:

  • Meals with mother;
  • Meals with father;
  • Participation in extracurricular activities;
  • School engagement;
  • Parents positive attitude toward community; and
  • Patents negative attitude toward community.

After adding the additional work pattern controls, the coefficient for parental aggravation fell to -0.212 and was no longer statistically significant at the .05 level.  Variables remaining statistically significant at the .05 level or better after applying both social and demographic controls and work pattern controls include:

  • Meals with mother;
  • Meals with father; and
  • Parents negative attitude toward community.

The meals with parent measures followed a noteworthy pattern.  As expected, for the meals with mother measure, with no controls there was no statistically significant difference between children in non-working poor families and children in working poor families  mirroring the bivariate results.  However, once basic demographic and social controls were added, the coefficient increased and became statistically significant at the .01 level.  This indicates that, after controlling for race/ethnicity, family structure, parental education, and parental age, children in non-working poor families ate more meals with their mothers than children in working poor families (an intuitively plausible result since higher work effort, other things equal, reduces time available for meals together).  The difference is reduced but remains statistically significant after adding the work pattern controls.

Even with no controls, fathers of children in working poor families were predicted to eat fewer meals with their children than the fathers of children in non-working poor families  mirroring the bivariate results.  When basic demographic controls were added, the difference widened, and, when the work pattern controls were added, the difference widened further.

Summary.  After controlling for race/ethnicity, family structure, parental education, and parental age, 10 of the 13 well-being measures between children in working poor families and children in non-working poor families that were significant in Tables 1-3 were also statistically significant in the multivariate analyses.  When controls for full-time/part-time work and full-year/part-year work were added, with a few key exceptions, the magnitude of the difference between these two groups of children were reduced and often lost their statistical significance.

View full report

Preview
Download

"report.pdf" (pdf, 134.43Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®