In this paper I study the causal relationship between fertility and female labor supply using census
data from 14 Latin American countries and the U.S. over the span of three decades (1980, 1990 and
2000). Parental preferences for a gender-balanced family (mixed-sex children) is exploited as a source
of exogenous variation in fertility. Although OLS estimates suggest a statistically significant negative
relationship in the 39 censuses used, instrumental variables approach fails to identify a causal effect
in most of them. The average effect of moving from a family with two children to more than two is
statistically zero for the group of compliers. Considering a pool of married women from Latin America
over the span of three decades, a negative causal effect is found. In any case, despite having a highly
accurate first-stage and indirect evidence consistent with the internal validity of the instrument, the
analysis of the quality of the instrument reveals a weak explanatory power of sibling sex composition on
fertility. The noisy and imprecise IV estimates for Latin America in the second-stage can be attributed
to the problem of weak instruments.