This paper offers empirical evidence on the impact of the expansion in health infrastructure of the 1990s upon child nutrition in Peru, as measured by the height for age z-score. Using a pooled sample of three rounds of the Peruvian DHS, I have controlled for biases in the allocation of public investments by using a district fixed effects model. The econometric analysis shows a positive effect of the expansion of the last decade in urban areas, but not in rural areas. Furthermore, the effect for urban children is highly non-linear and has a pro-poor bias, in the sense that the estimated effect is larger for children of less educated mothers. These findings support the idea that reducing distance and waiting time barriers is necessary to improve child health and nutrition in developing countries, but that we need more explicitly inclusive policies to improve the health of the rural poor, especially indigenous groups, that are caught in this type of poverty trap. © 2004 Elsevier B.V. All rights reserved.