The migration literature typically assumes that the migration of a household member is not associated with further variations in co-residence choices. We rely on a Mexican panel survey to provide novel evidence on the correlation between the occurrence of an international migration episode and changes in household composition. Migrant households have a higher probability of receiving a new member within one year around the migration episode. Attrition is significantly higher among migrant households, and we provide evidence that this is partly due to the dissolution of the household of origin of the migrant. The endogeneity of co-residence choices induces an undercount of migration episodes, as shown with data from the 2000 Census. This has implications for the analysis of migrant selection and of the effects on the individuals left behind. Dealing with these analytical challenges requires an approach to data collection that is less dependent on variations in household composition.