Comments

Gerardo Esquivel: Hildebrandt and McKenzie study the impact of migration on child health indicators in rural Mexico. They use a 1997 nationally representative survey that contains information on both household members' past migration decisions and children's health outcomes. To address the likely endogeneity of migration, they use historic migration patterns as instruments for current migration. The paper finds that children in households with migrants have lower infant mortality rates and a lower probability of being underweight than do children in nonmigrant households. It also finds that one of the channels through which migration seems to influence children's health is an increase in their mothers' health knowledge, as proxied by her knowledge of ten different contraceptive methods. Somewhat paradoxically, the paper also finds that children in households with migrants have a lower probability of being breast-fed, of having received all relevant vaccinations, and of having visited a doctor at least once during their first year of life.

The issues addressed in this paper are important not just for Mexico, but for any country with substantial emigration. Understanding the effects of migration on sending countries is at least as important as studying its effects on receiving ones. The former line of analysis has received little attention, however, and this paper helps to fill the gap. The authors also develop an estimation strategy to address the endogeneity problem that pervades this type of study. In fact, they suggest that the difference between their results and those in a couple of sociological papers arises precisely from the correction for endogeneity and self-selection into migration. The authors are also careful in demonstrating the robustness of their conclusions.

In general, I like the paper very much and find its approach quite appealing. At the same time, I am not entirely comfortable with the authors' position on what appears to be conflicting evidence on the effect of migration on child health. For example, how do they explain the fact that that migration increases mothers' health knowledge but that children in migrant households are less likely to have a complete set of vaccinations and less likely to visit a doctor in their first year of life? These results seem odd. I hope the subject will be addressed in future research, so that the literature can eventually reach more conclusive results on the net impact of migration on child health.

I am also somewhat skeptical about the magnitude of some of the estimates presented, particularly the estimates of the effect of migration on infant mortality rates. The authors conclude that children in a household with migrants are between 3.0 and 4.5 percent less likely to die in their first year of life, depending on the estimation method, than children born in a household without a migrant member. These estimated effects seem high given that a child's unconditional probability of dying in his or her first year of life is only about 2.7 percent and that the marginal effects on a probit estimation are in absolute and not in relative terms. This problem ultimately reflects the fact that the effects are calculated assuming a discrete change in the instrumented variable (in this case, the migrant household variable; see the note in table 5). However, when this variable (or any unbalanced binary variable—that is, one with many more zeros than ones) is instrumented, the range in which the estimated variable moves is shortened significantly owing to the nature of the first-step estimation when we have an endogenous binary variable. This problem is not easily solved, and it has been overlooked in the empirical and theoretical literature on instrumental variables with an unbalanced endogenous binary variable. One should therefore be careful in interpreting this type of result. All in all, I think Hildebrandt and McKenzie make a very important contribution to the growing literature on the impact of migration on the sending economy. I praise them for being among the first authors to engage in such an important line of research and for their methodological contributions, which should definitely be taken into account in...