Marco Gonzalez-Navarro and Climent Quintana-Domeque have an interesting paper which studies the impact of a randomized allocation of residential non-arterial streets (8-15 m) for asphalting in Acayucan municipality of Mexico,
Within two years of the intervention, households whose streets were finally paved, and were present both before and after its implementation, increase their consumption of durable goods and acquire more motor vehicles. These effects are driven in part by street pavement boosting housing wealth, which fuels a rise in collateralized credit use. The effects appear to be present for both the relatively poor and the relatively rich households in our sample... Our main conclusion is that provision of street pavement reduces poverty, at least as measured by the increases in the acquisition of durable goods, motorized vehicles, and property values.
Finally, we estimate a lower bound for the total benefit to cost ratio of street pavement at around 1.09, which justifies serious consideration of cost-sharing mechanisms in the provision of localized benefit public goods. In the absence of credit constraints, property taxes triggered by public goods that increase land value could be used as a funding device for localized impact urban infrastructure, such as Colombia’s urban improvements tax or U.S. road paving special assessment districts.
The acute financial constraints faced by municipalities in developing countries prevent them from making these basic investments whose beneficial effects are undoubted. In the circumstances, such resource allocation - which streets get selected on priority - becomes a political decision, with its set of perverse incentives. This goes back to the irony that Donald Shoup posed evocatively, “Why is it so difficult to finance public infrastructure that increases the value of the serviced land by much more than the cost of the infrastructure itself?” In this context, the adoption of value capture finance strategies like tax increment financing and impact fees assume significance.