Estimation Results

4.1 Labor Force Participation of Mothers

Both the endogeneity tests using either the two-stage probit or the FIML all show that the number of children using both boys and girls or same sex⁴ in the first two births as instruments did not show significance. Table 4 [ PDF 128.2KB | 12 pages ] shows, the error in number of children did turn out to be significant with z-value of 1.4. In addition, the test of the correlation coefficient from the FIML estimation for both the mother and father labor force participation equation also turned out to be insignificant with chi-square value of 0.2 (Table 5 [ PDF 128.2KB | 12 pages ]). Both of these imply that for this particular data set the endogeneity of the number of children is not established. This also lends support to the validity of using simple probit estimates. Subsequent discussion will then focus on the simple probit estimates.

The simple probit estimates of the determinants of the labor force participation of mothers are given in Table 6 [ PDF 128.2KB | 12 pages ]. The labor force participation of the mothers are expected to decline by 0.92⁵ percentage points with each additional child. This slightly rises to 0.96 percentage points when only unpaid work is excluded. Another noteworthy result relative to the impact of children on the labor force participation of their mothers is the impact of the presence of children below normal school age – 0 to 5 years old. Table 5 [ PDF 128.2KB | 12 pages ] shows that the presence of young children below normal school age reduces the probability of their mother working by a considerably higher 7.8 percentage points when all types of work are considered and 5.7 percentage points when unpaid work is excluded. This confirms many of the results in previous studies on the negative impact of children on the labor force participation of mothers.

The other significant determinant variables of mother’s labor force participation are her age, education, wage income of the father and unemployment rate. The age of the mother was entered as a quadratic to capture non-linear effects. The signs of the coefficients confirm the expectation that labor force participation of mothers rises at a declining rate with age. Education is a positive determinant as found in many other studies. The higher the wage income of the father the lesser is the likelihood that a mother would be working. Interestingly, the estimates shows that a thousand increase in the father’s wage income per capita⁶ would have an equivalent depressing effect as an additional child. A higher unemployment rate discourages mothers from looking for work lending support to the “discouraged worker” hypothesis. Contrary to expectation, nonwage income per capita of the household is a significant positive determinant of the labor force participation of mother for all types of work although it has the expected negative sign in the paid work equation.

To determine the differential impact of children on the labor force participation across income classes, the number of children variable was interacted with the per capita income quintile dummy variables. The result of the estimation is given in Table 7 [ PDF 128.2KB | 12 pages ]. For all types of work, the interaction variables are not significant except for the top two quintiles. For paid work, however, all the interaction terms are significant. Table 8 [ PDF 128.2KB | 12 pages ] provides the summary of the impact on the labor force participation of mothers by per capita income quintiles, expressed as a percentage of the recorded participation rates. For all types of work, mothers from the bottom three quintiles will reduce the proportion working by an average of over 2 percent for each additional child. For the top two quintiles, however, the impact is positive implying that more of them will work with an additional child. This could even reach as high as about 7% for the richest quintile. In the case of paid work, the pattern is similar except that the impacts are much larger in magnitude. About a 6% decline for each additional child for mothers in the bottom quintile and more than 8% increase for each child for mothers in richer quintiles. This differentiated impact undoubtedly provides a richer view of the impacts than just the average impact.

4.2 Labor Force Participation of Fathers

Similar to the results in the mothers’ labor force participation equation, the test results for endogeneity of the number of children in the fathers’ labor force participation equation also had insignificant results. The coefficient of the estimated error term in the first stage regression did not turn out to be significant in the second stage labor force participation equation with z value of 0.16 (Table 9 [ PDF 128.2KB | 12 pages ]). Similarly, the test of the correlation coefficient using FIML also turned out to be insignificant with a chi-square value of .03 (Table 5 [ PDF 128.2KB | 12 pages ]). These lend support to the validity of using simple probit results.

The results for the labor force participation of fathers show that on the average the number of children does not affect their labor force participation (Table 10 [ PDF 128.2KB | 12 pages ]). It is negative but not significant. This means that fathers, on the average, do not try to find work when a child is added to the family.

Similar variables that are significant in the mother’s labor force participation equation are also significant in the father’s labor force participation equation. Age of the father has a rising at a declining rate effect. One surprising results is the negative and significant coefficient for the years of education of the father. Perhaps, this may mean that highly educated fathers are earning more from other sources. Non-wage income is not a significant determinant. Like for mothers, the discouraged worker hypothesis also works for fathers, i.e., with higher unemployment they tend not to look for work all other things equal.

Again to determine the differential impact across income classes, the number of children variable was interacted with the per capita income quintile dummy variables. The estimation results are also shown in last four columns of Table 10 [ PDF 128.2KB | 12 pages ]. All the coefficients of the interaction terms are significant. Since the base category is not significant it will be considered as zero. The summary of the impact expressed in terms of the recorded proportion of fathers working is also provided in Table 8 [ PDF 128.2KB | 12 pages ]. For the poorest quintile there would be no effect. The impacts for the upper income quintiles are all positive with 0.3, 0.6, 0.4, and 1.2 for the 2nd, 3rd, 4th and 5th quintiles, respectively. Thus, there is slight positive effect for the labor force participation of fathers in the upper income quintiles.

4.3 Earnings of Mothers

The endogeneity test using two-stage tobit shows that the number of children is not endogenous in the earnings equation for this particular data set as is seen in the labor force equation. The coefficient of the estimated first-stage error term is not significant with a t value of 0.63 (Table 11 [ PDF 128.2KB | 12 pages ]) lending support to the validity of using ordinary tobit estimates. Thus, the subsequent discussion will refer only to the ordinary tobit results. Table 12 [ PDF 128.2KB | 12 pages ] shows the results of the Tobit and OLS estimates for the earnings of mothers. The number of children is found to negatively affect the earnings of mothers on the average. But when one looks at the coefficient of the interaction terms with per capita income, the negative impact is only for the bottom two quintiles⁷. The upper three quintiles have positive impacts and this is roughly consistent with the results for labor force participation. It is easier to look at the impact as a percentage of recorded incomes and in absolute value.⁸ These are shown in Table 13 [ PDF 128.2KB | 12 pages ]. The average effect is about a 5% decline in income or about 1 thousand pesos from the six-month earnings per additional child. For the bottom two quintiles, the impact is about -13% for the poorest and -7% for the lower middle quintile per additional child. These translate to about a reduction of about 700 and 600 pesos to the semesters wage income, respectively. For the top three quintiles, the impacts are positive: 2%, 15% and 33% for the middle, upper middle and richest quintile, respectively. This means an addition of 360, 6,200 and 25,736 pesos to the semester wage income for the corresponding quintiles, respectively, per additional child.

All the other variables, age, education, residence in urban areas and regional dummies are significant in determining the wage income of mothers. Wage income rises with age in a decreasing manner. Education positively affects earnings. There is on average higher earning in urban areas. Except for ARMM, which is unexpected, mothers in all other regions earn lower than those in the national capital region.

4.4 Earnings of Fathers

Again the endogeneity test using the two-stage tobit did not yield significant results. The coefficient of the estimated first-stage error term is not significant with a t value of 0.83 (Table 14 [ PDF 128.2KB | 12 pages ]) lending support to the validity of using the ordinary tobit results. This is what we use in subsequent discussions. The result shows that on average, the earnings of fathers are positively affected by the presence of children (Table 15 [ PDF 128.2KB | 12 pages ]). This is in contrast to the impact on labor force participation that showed no significant affects. Perhaps fathers are more serious in finding paying or higher paying jobs rather than just any job with additional children. The impact as a percentage of recorded income and in absolute value is also given in Table 13 [ PDF 128.2KB | 12 pages ]. There is an average increase of about 1% in income of fathers or an addition of about 233 pesos to the semester wage income. The interaction terms between the number of children and per capita income quintile dummies, also all turned out to be significant. Considering all of these, the negative impact remains for the bottom quintile with about a 6% decline in wage income or about a 76 pesos reduction in semester income per additional child. For the lower middle up to the richest quintile, the effect is positive from 93 to 25,538 pesos addition to the semester wage income for the lower middle to the richest quintile per additional child. Judging from these numbers, even though the impact is positive, this is hardly enough to pay for the marginal increase in expenditure due to the new child, except perhaps for the richest quintile.

The other variables have a similar performance with those in the mothers’ earnings equation. Earnings rise with age at a declining rate. Education positively affects income. There is a higher average earnings of fathers in urban areas. Earnings of fathers in the national capital region are also higher than those in the other regions, without exception.

Download this Discussion Paper [ PDF 268.9KB| 27 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.