Urban Public Works in Spatial Equilibrium: Experimental Evidence from Ethiopia

Abstract

Anti-poverty programs affect not only beneficiaries, but also the entire economy, through spillover effects which are often difficult to identify and quantify. This paper evaluates an urban public works program, Ethiopia's UPNSP, which provides employment on local infrastructure projects to poor households. For identification, we use the random roll-out of the program across neighborhoods of Addis-Ababa. We develop a spatial equilibrium model and leverage unique data on local amenities and city-wide commuting flows to account for spillover effects. We first present reduced-form evidence that public works partially crowded-out private sector employment, increased private sector wages and improved local amenities. We then use a model-based measure of exposure to estimate the wage effects of the program, which were felt across the city: wages increased by 20% in program neighborhoods and 10% in other neighborhoods. Finally, we compute the welfare gains to the poor once the program was completely rolled-out across the city: 66% come from rising private sector wages, 12% from improvements in local amenities, and only 21% from public employment. These results suggest that welfare gains are five times larger after taking spillover effects into account. (joint with Simon Franklin, Girum Abebe and Carolina Mejia-Mantilla)