Researchers have developed a new method called $\ell$BISG (list-powered Bayesian Improved Surname Geocoding) to probabilistically predict race and ethnicity using name lists. This approach overcomes the limitations of existing methods that rely on group frequencies for specific names, which are often unavailable, especially outside the United States. $\ell$BISG can utilize name lists generated by large language models (LLMs) and can even function without geographic prevalence data under certain conditions. Empirical studies using U.S. voter files, historical census data, and a Lebanese voter registry demonstrated that LLM-generated lists produce accurate and well-calibrated predictions, extending the reach of probabilistic race and ethnicity estimation. AI
IMPACT Enables more accurate and widespread probabilistic race and ethnicity prediction, particularly in regions lacking traditional frequency data.
RANK_REASON The cluster contains an academic paper detailing a new methodology for probabilistic race and ethnicity prediction. [lever_c_demoted from research: ic=1 ai=0.7]
- 1900 U.S. Census
- Bayesian Improved Surname Geocoding
- \ell BISG
- large language models
- Lebanese voter registry
- Noah Dasanaike
- United States Census Bureau
- U.S. voter files
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →