The hidden cost of segregation: Unequal energy burdens and causal evidence

Numerous researchers have conducted correlational analysis to identify factors contributing to household energy poverty. While proposing various determinants, most have struggled to convincingly establish causal relationships. This paper presents a comprehensive database created by merging mortgage application and denial rates with housing unit structural characteristics, energy-related expenses, and socioeconomic data describing utility consumers in Atlanta, Georgia (one of the most energy-burdened cities in the United States). Its high level of geographic resolution (the census tract) spanning multiple years enables an innovative causal identification methodology. In combination with various sensitivity analyses, we estimate the causal effect of neighborhood racial composition on energy burden using mortgage denial rates as instrument. We find that an approximate 11 percentage-point increase in the Black share of a neighborhood corresponds to a 1 percentage-point higher household energy burden (defined as utility costs relative to household income), equivalent to an average annual energy cost increase of 785. Robustness tests include replication in a second major city (San Diego, California), single-year and panel data models, alternative definitions of energy burden, reduced-form analysis using historical redlining, and a sensitivity analysis using individual-level data (to address potential ecological fallacy concerns) – all reinforce our findings on the hidden cost of segregation.