This paper evaluates a large urban public works program randomly rolled out across neighborhoods of Addis Ababa, Ethiopia. We find the program increased public employment and reduced private labor supply among beneficiaries and improved local amenities in treated locations. We then combine a spatial equilibrium model and unique commuting data to estimate the spillover effects of the program on private sector wages across neighborhoods: under full program rollout, wages increased by 18.6 percent. Using our model, we show that welfare gains to the poor are six times larger when we include the indirect effects on private wages and local amenities.