Space matters not only because of the transportation costs it imposes on the economy but also because it can serve as an effective instrument to control pollution damages. Previous models of pollution either disregard space altogether or presume a predetermined separation between polluters and pollutees, usually into a CBD where the polluting industry is located and a residential ring where the city's laborers reside. All possible location combinations of housing and industry are considered in this study. The results demonstrate that the management of pollution must recognize the trade-off between two cost components: pollution costs and transportation costs. This trade-off along with the non-convexity inherent in spatial models results in multiple local optima. Negligible commuting costs combined with pollution emissions bearing ill effects at a rate declining with distance leads to an allocation with one industrial zone and one residential zone. As commuting costs increase, the optimal allocation passes through an endogenously determined series of increasing thresholds. Each time a threshold is crossed the number of zones of each type increases until the internal solution is reached after the final threshold has been crossed. In the internal solution, there is no commuting, and housing and industry assume adjacent locations. In such an economy, Pigouvian taxes are generally inefficient. Instead, the efficient tax is a per unit land tax equal to the additional damages contributed by that land unit's pollution.

Previous analysis of pollution presumes separation between pollutees and polluters. In the model developed here, all possible combinations of housing and industry locations are allowed. In the resulting world of nonconvexities and multiple local optimum, Pigouvian taxes are generally insufficient. Accordingly, our analysis considers a wide range of policy instruments extending from Pigouvian taxes, double taxes, to zoning regulations. The results demonstrate that the management of pollution requires the recognition of two separate regimes determined by the type of convexity or concavity of the pollution dispersion function. When this function is convex, the optimal solution requires no zoning of housing. When the dispersion function is concave in emissions, the optimal allocation implies zoning into industrial and residential zones and, in some circumstances, taxes equal to total damages. To achieve effective management under limited information regarding the pollution dispersion function, it is argued that zoning restrictions cab be determined by trial and error through observation of changes in land rents.