Abstract
Regional policies often affect untreated but spatially exposed units, making standard difference-indifferences comparisons difficult to interpret. This paper develops a treatment-exposure framework for regional policy evaluation with spatial spillovers. It shows that two-way fixed effects estimators that omit exposure conflate the direct treatment effect with spillover and treated-exposure components whose signs depend on the exposure design. A decomposed specification separates direct effects, spillovers on untreated exposed units, and reinforcement among treated exposed units, provided that the spatial weights matrix is specified as a causal exposure mapping before estimation. The paper translates these conditions into an identification protocol for selecting, diagnosing, and challenging exposure mappings. An application to London's Ultra Low Emission Zone estimates a direct NO 2 reduction of 6.6 percent and an implied reduction of 6.2 percent at the mean positive exposure level among untreated exposed units. Spatially bounded evaluations restricted to directly treated units may understate the affected population and misstate the costs and benefits of the intervention.

This publication has 31 references indexed in Scilit: