Is infrastructure construction an effective way to boost employment in distressed local labor markets? I use new geographically-detailed data on highway construction funded by the American Recovery and Recovery Act to study the relationship between construction work and local employment growth. I show that the method for allocating funds across space facilitates a plausible selection-on-observables strategy. However, I find a precisely-estimated zero effect of spending on road construction employment–or other employment–in the locale of the construction site. Reported data on vendors reveal this is because the majority of contractors–selected by competitive bidding–commute from other local labor markets. I also find no robust effect in the locales of the contractors’ offices, but argue that this source of variation does not capture an economically meaningful local demand shock. I conclude that infrastructure construction is not effective as a way to stimulate local labor markets in the short-run so long as projects are allocated by competitive bidding.