3.2 Results – Census Data
The estimation results for the samples of children whose mothers are working are presented in Table 3, and those for children whose mothers are not working are presented in Table 4. In Table 5, we present the estimation results for the pooled 2001 and 2006 Census datasets including both working and nonworking mothers.
Column 1, Table 3 shows the estimation results when express schooling is used as the dependent variable for the 2001 Census dataset. It shows that employing a livein Filipino FDW increases the probability of express schooling for children under their care by 3.7%, which has a Pvalue of 0.000. Employing an Indonesian FDW also increases the probability of express schooling for children under their care by 3.3%, which has a Pvalue of 0.007. As expected, household income, being a female child, parents’ education, and the length of the mother’s residency in Hong Kong all increase the school child’s probability of express schooling. On the other hand, the greater the number of children in the household and mother’s age decrease the children’s chances of express schooling. All the coefficients in Column 1, Table 3 are precisely estimated except for the number of elderly in the household and the father’s age, both of which are not statistically significant at the conventional levels. Column 2, Table 3 shows the estimation results of Eq. (2) for the 2006 Census dataset when express schooling is used as the dependent variable. Again, the estimated effect of employing a Filipino FDW or an Indonesian FDW on children’s probability of achieving express schooling is positive and highly significant, being 3.8% and 3.2% respectively, similar to the outcomes of the 2001 Census dataset. Moreover, as in Column 1, the estimated effects of other control variables in Column 2 are consistent with our prior expectation, even though household income, being a female child, the number of elderly, and mother’s education are no longer estimated precisely.
Columns 3 and 4 of Table 3 present estimation results of Eq. (2) when late schooling is used as the dependent variable. Column 3 shows the results for the 2001 Census dataset and Column 4 for 2006 Census dataset. Both Columns 3 and 4 show that employing a livein Filipino FDW reduces the children’s chances of late schooling, but the estimated effect is only marginally significant in Column 3 and nonsignificant in Column 4. For Indonesian FDWs, both estimates in Columns 3 and 4 show little effect on the children’s probability of late schooling. Estimated coefficients in both Columns 3 and 4 relating to the child being female, the number of elderly in the household, parents’ education, parents’ age, and the length of mother’s residency in Hong Kong are all consistent with our prior expectation. An exception is that the estimates for the number of children in the household show negative signs in Columns 3 and 4, which is statistically significant, indicating that the greater the number of siblings in the family reduces the chances of late schooling for the children.
Table 4 presents the estimation results of Eq. (2) for the sample of nonworking mothers. Mothers who are not in the workforce are able to spend relatively more time with their children and thus the effects of FDWs on children are expected to be weaker. When express schooling is used as the dependent variable in Columns 1 and 2, both the 2001 and 2006 Census datasets show that employing a livein Filipino or Indonesian FDW is estimated to increase the probability of children’s express schooling. However, only the 2006 estimates are statistically significant, while those of the 2001 are not. When late schooling is used as the dependent variable in Columns 3 and 4, both the 2001 and 2006 datasets show little effects of (Filipino or Indonesian) FDWs on children’s late schooling. Other estimates in Table 4 are similar to those of Table 3, except for the length of mother’s residency in Hong Kong which is estimated to reduce children’s express schooling in Table 4. This result seems to suggest that mothers who arrive and continue to live in Hong Kong but are unable to find work have a negative effect on children’s express schooling.
Table 5 shows the estimation results when both 2001 and 2006 Census datasets are pooled for estimating Eq. (2). Pooling the two datasets increases the precision of estimates by increasing the number of observations. A 2001 dummy variable is added to the regression to control for the year effect. Column 1, Table 5 shows that for children whose mothers are working, a Filipino FDW increases their probability of express schooling by 6.2% and an Indonesian FDW by 3.7%, both of which have a Pvalue less than the 1% level. Column 2 shows that for children whose mothers are working, a Filipino FDW reduces their probability of late schooling by 3.3%, which is statistically significant at less than 1% level. In contrast, the estimated effect of Indonesian FDWs’ on children’s probability of late schooling is nonsignificant. Estimates in Columns 3 and 4 show that for children whose mothers are not working, employing either a Filipino or Indonesian FDW increases the children’s probability of express schooling, but has little effect on the children’s probability of late schooling. The 2001 dummy indicates that there is a significant increase in both express and late schooling in 2006 compared to 2001 for working and nonworking mothers. Other estimates in Table 5 are largely similar to those in Tables 3 and 4 and thus are not reported in the interests of space.
In sum, the estimation results based on the Census datasets shown in Tables 3, 4 and 5 indicate that employing a livein Filipino or Indonesian FDW improves the probability of express school for children under their care by as much as 6%. As expected, this estimated FDW effect is stronger and more statistically significant for the sample of working mothers than for nonworking mothers. Also, Indonesian FDWs generally show a weaker estimated effect on the children’s probability of express schooling compared to that of the Filipino FDWs. If we use late schooling as the dependent variable, the FDW effect is harder to detect but is still clearly reflected in the case of children who are cared for by a Filipino FDW while their mothers are working. The overall results from the Census datasets thus show evidence of FDWs contributing to children’s schooling achievement, which is not explained by household income, family background and children’s characteristics.

Share with your friends: 