Recent empirical studies document that knowledge spillovers attenuate and industry localization decays with distance. It is thus imperative to detect localization accurately especially at short distances. We propose a new approach to testing for localization that corrects the first-order bias at and near the boundary in existing methods while retaining all desirable properties at interior points. Employing the NBER U.S. Patent Citations Data File, we illustrate the performance of our localization measure based on local linear density estimators. Our results suggest that the existing kernel density methods and regression approaches can be substantially biased at short distances.