Adam Storeygard, Tufts University Transport costs and city growth across sub-Saharan Africa: short- and long-run estimates Adam Storeygard, Tufts University
Challenges of impact evaluation General Counterfactual is unobserved: what would have happened in absence of program Treated region may be different from others for many reasons: Omitted variables Reverse causality Project built to accommodate a region expected to grow Project built to help a region expected to lag Large infrastructure projects Spatial scale of impacts unclear (potentially large) Harder to find a good counterfactual/control group
Project 1: interannual transport costs and city growth Key ideas Many African countries have a large coastal city Largest domestic market Gateway to rest of world Other cities need to trade with it Transport cost to it is critical Transport cost = distance * cost/distance Proxy for cost/distance with oil price Proxy for city economic activity with nighttime lights Limited alternative city-specific outcome data for a large number of cities
oil price increase of $72/barrel (~2002-2008) Results oil price increase of $72/barrel (~2002-2008) Cities near port grow by 7 percent relative to cities 500 km further away Suggests elasticity of city economic activity with respect to transport costs of -0.25 at 500 km If transport costs by 10%, city economic activity by 2.5 % (relative to other cities in same country) Cities connected to port by paved roads are chiefly affected by transport costs to the port Less connected cities more affected by connections to secondary cities
Project 2: road-building and long run city growth (w/Jedwab) Key ideas Digitized all Michelin road maps for sub-Saharan Africa 1961-2014 Define a city’s market access as the ease with which its firms/residents can travel to all other cities (weighted by their population) Better roads increase market access Better roads near a city might be endogenous Instrument for a city’s overall Δ(market access) with Δ(market access) based only on roads built far away Population as long-run outcome
Elasticity of 5% to 20% depending on instrument (radius of exclusion) Preliminary results Elasticity of 5% to 20% depending on instrument (radius of exclusion) Larger than naïve OLS estimates of 3-4% Doubling market access increases population by 5-20% Some evidence it is urbanization, not displacement from other (nearby) cities Effect spread over 30 years Suggestion of larger effects for smaller, more remote cities
Evaluation of transport projects requires careful consideration of Conclusions Some evidence that transport costs matter for the growth of African cities Evaluation of transport projects requires careful consideration of Relevant counterfactuals Omitted variables Reverse causality Some experiments, quasi-experiments possible Random paving in Mexico Policy rules in India, Sierra Leone