Constructing Confidence Intervals for BIFSG Disparity Estimates

Abstract

This paper describes a method to construct confidence intervals for estimates of disparities in outcomes between racial/ethnic groups when race and ethnicity are estimated using Bayesian Improved First Name and Surname Geocoding (BIFSG). Standard errors produced by regression packages systematically understate the uncertainty in these estimates because they fail to incorporate additional noise arising from the estimation of race and ethnicity. This can lead to false confidence in assessing both the direction and magnitude of disparities. Our approach uses a bootstrapping procedure, and reflects substantially more of the uncertainty introduced when using BIFSG. However, this procedure does not correct for any statistical bias in disparity estimates associated with using BIFSG.

Keywords: BIFSG, Race, Ethnicity, Confidence Interval, Inference, Standard Error, Tax

JEL Classification: H2, C11, C81