Kyle Butts1
1University of Colorado: Boulder
I formalize a commonly-used estimator for the effects of spatially-targeted treatment with geocoded microdata. This estimator compares units immediately next to treatment to units slightly further away. I introduce intuitive identifying assumptions for the average treatment effect among affected units and illustrate problems when these assumptions fail. I propose a new method that allows for nonparametric estimation following methods introduced in Cattaneo et al. (2019) that allows estimation without requiring knowledge of exactly how far treatment effects are experienced. Since treatment effects can change with distance, the proposed estimator improves estimation by estimating a treatment effect curve.
Figure 1: Rings Method
figure-example_problems.R
Figure 2: Example of Problems with Ad-Hoc Ring Selection
figure-example_problems.R
Figure 3: Price Gradient of Distance from Offender
analysis-linden_rockoff.R
Figure 4: Effects of Offender Arrival on Home Prices (Linden and Rockoff 2008)
analysis-linden_rockoff.Rhelper-nonparametric_rings_estimator.Rhelper-parametric_rings_estimator.Rhelper-plot_rings.R
Table 1: Monte Carlo Simulations
analysis-simulations.R
@article{butts2023jue,
title={JUE Insight: Difference-in-differences with geocoded microdata},
author={Butts, Kyle},
journal={Journal of Urban Economics},
volume={133},
pages={103493},
year={2023},
publisher={Elsevier}
}