Versión de HTML Básico
Andrea Molina Vera
Analítika, Revista de análisis estadístico, (2015), Vol. 9
14
•
Worked for pay outside the home: equals 1 when the mother’s job is paid and outside
the home.
The dimensions that can be explored with these definitions are: location of job in relation
to home, remuneration and labor participation.
4 Main Results
Following Angrist and Evans, table 3 shows estimates of the impact of child sex and sexmix
on fertility, where 48.9% of all women have one girl and 51.1% have one boy at the first birth.
The fraction of women with at least one child and who had a second child, conditional on
the sex of the first child is 67% in both cases. This presents evidence that there is no impact
of the sex of the firstborn on fertility. It is important to mention that, in married women,
there are differences conditionals on the sex of the first child but this difference disappears
when including controls in the regression at the first stage.
For the second analysis, table 3 presents the fraction of women who have a third child
conditional on the sex composition of the first two children, where 46.4% of women with
one boy and one girl have a third child, compared to 49.9% for women with two girls or two
boys. That is a significant difference of 3.6 percentage points. These results are confirmed
in table 4 which shows the first stage of the instrument for all and for married women,
including controls, and results for the other possible instrument (two boys and two girls).
The difference of 3.6 and 3.8 percentage points for all and for married women found here
means that Ecuadorian women with two children of the same sex are 3.6 and 3.8 percentage
points more likely to have a third child than mothers of one boy and one girl. For the United
States in 1980 this difference was 6 percentage points for all women (Angrist and Evans).
Cruces and Galiani found a difference of 3.5 and 3.2 percentage points (subsample of all
women) for Argentina and Mexico, respectively. On the other hand, the instrument of two
boys and two girls is also significant for explaining fertility.
To check a random assignment of Same-sex instrument, table 5 compares demographic
characteristics of the mother among those who had a composition of same-sex (treated group)
and mixed-sex (control) sibling compositions. This table includes the following variables:
age of woman, age at first birth, indigenous ethnicity, years of education and residence area.
None of these variables presents significant difference
For 2SLS estimates the control variables are: age of women, age at first birth, a dummy
variable to indicate the sex of first and second children, a dummy variable for ethnic identifi-
cation and a dummy for urban area. For the cases of two boys and two girls as instrument, the
covariates exclude the sex of the second children. Table 6 shows that OLS estimates present
a negative and significant impact for all measurements of work and all groups of women.
The 2SLS estimates indicate there is an impact on labor supply when moving from 2 to 3
children for all women grouped in two work measurements (working outside from home and
8