Estimating Racial Disparities When Race is Not Observed

Estimating racial disparities without access to individual-level racial information is a common challenge in economic and policy settings. We develop a statistical method that relaxes the strong independence assumption of common race imputation approaches like Bayesian-Improved Surname Geocoding (BISG). Our identification assumption is that surname is conditionally independent of the outcome given (unobserved) race, residence location, and other observed characteristics. The proposed approach reduces error by up to 84% relative to BISG when estimating racial differences in political party registration. In our application, we estimate racial differences in who benefits from the home mortgage interest deduction using individual-level tax data from the U.S. Internal Revenue Service. Our analysis reveals that many fewer Black and Hispanic filers claim the HMID than White and Asian filers. We also find that the racial gaps in homeownership rates alone cannot explain this disparity. Supplementary materials for this article are available online, including a standardized description of the materials available for reproducing the work.

在经济学与政策研究场景中，无法获取个体层面种族信息时开展种族差异估算，是一类常见的研究难题。我们提出了一种统计方法，可放宽贝叶斯改进姓氏地理编码法（Bayesian-Improved Surname Geocoding，BISG）等常见种族插补方法所采用的强独立性假设。我们的识别假设为：在给定（未观测到的）种族、居住区位与其他观测特征的条件下，姓氏与结果变量条件独立。在估算政党登记的种族差异时，所提方法相较BISG可将误差最高降低84%。在本研究的应用场景中，我们借助美国国税局（Internal Revenue Service，IRS）提供的个体层面税务数据，估算了享受住房抵押贷款利息扣除（Home Mortgage Interest Deduction，HMID）的群体所存在的种族差异。分析结果显示，申报HMID的黑人和西班牙裔纳税人数量，远低于白人与亚裔纳税人。我们还发现，仅以住房拥有率的种族差距，无法解释这一差异。本文的补充材料可在线获取，其中包含可复现本研究的相关材料的标准化说明。