Simulating confidence for the Ellison–Glaeser index

The Ellison and Glaeser (1997) index is an unbiased statistic of industrial localization. Though the expected value of the index is known, ad hoc thresholds are used to interpret the extent of localization. We improve the interpretation of the index by simulating confidence intervals that a practitioner may use for a statistical test. In the data, we find cases whose index value is above the ad hoc threshold that are not statistically significant. We find many cases below the ad hoc threshold that are statistically significant. Our simulation program is freely available and is customizable for specific applications.