Measuring Descriptive Representation at Scale: Methods for Predicting the Race and Ethnicity of Public Officials

Abstract

Political scientists have increasingly relied on Bayesian Improved Surname Geocoding (BISG) to predict individual-level race and ethnicity using surnames and geographic information. However, these predictions are prone to systematic measurement error when geographic variables are included as independent variables, leading to bias and an increased likelihood of false positives. We compare five estimation methods, including approaches that use names, geography, and images. We find that a hybrid approach that combines surname-based Bayesian estimation with the use of publicly available images in a convolutional neural network not only reduces bias in downstream analyses, but also improves predictive accuracy in a sample of over 16,000 local elected officials. We conclude with a discussion of ethical considerations and describe domains where the hybrid approach might be especially suitable.

Rising Tides or Political Ripcurrents? Gentrification and Minority Representation in 166 Cities

“…Better understanding how councilors respond to the increasingly ubiquitous phenomenon of gentrification could yield important insights about the changing nature of urban politics in the twenty-first century (Carr and Zeigler 1990). demographic information of an individual, has been proven to be quite effective (Shah and Davis 2017;Lee and Velez 2023). 5.…”