Race and ethnicity data for first, middle, and surnames

Area of Science:

• Demographic research
• Computational social science
• Public health data analysis

Background:

• Accurate imputation of race and ethnicity is vital for understanding population demographics and health disparities.
• Existing name-based imputation resources are limited in scope and coverage.
• Self-reported racial data collection is not always feasible or available in large datasets.

Purpose of the Study:

• To compile and release the largest publicly available dictionaries of first, middle, and surnames for race and ethnicity imputation.
• To provide a comprehensive resource for methods like Bayesian Improved Surname Geocoding (BISG).
• To facilitate accurate demographic analysis in the absence of self-reported data.

Main Methods:

• Compiled dictionaries from voter files of six U.S. Southern States with self-reported racial data.
• Included 136,000 first names, 125,000 middle names, and 338,000 surnames.
• Calculated probabilities of race/ethnicity given a name (ℙ(race|name)) and name given race/ethnicity (ℙ(name|race)).

Main Results:

• Created the most extensive name dictionaries for race and ethnicity imputation currently available.
• Data covers five mutually exclusive groups: White, Black, Hispanic, Asian, and Other.
• Provided conditional probabilities for each name, enabling robust imputation.

Conclusions:

• The released name dictionaries represent a significant advancement for demographic imputation.
• These resources can improve the accuracy of race and ethnicity estimation in large-scale data analysis.
• Facilitates research requiring demographic data where direct reporting is not possible.