Determinants of U.S. residential energy consumption at national and state levels: Policy implications

Abstract

Literature has extensively studied the effects of physical and environmental parameters on building energy consumption through physics-based simulations. However, energy use of the building sector at state and national levels is influenced by more complex factors that vary spatially and temporally and include sociodemographic, socioeconomic, physical, climatic, and microclimatic parameters.

This study identifies key determinants of U.S. residential energy consumption at national and state levels using the 2020 Residential Energy Consumption Survey (RECS) data and examines whether they increase or decrease energy use. This research uses Machine learning algorithms for energy modeling and SHAP (SHapley Additive exPlanations) sensitivity analysis to explain the contribution of each feature to the model. Top determinants influencing national energy include using electricity for space and water heating, Heating Degree Days, and energy-consuming areas. Using electricity for heating can significantly reduce on-site residential energy consumption. Given that electricity is the second most common heating source in American homes after natural gas, these findings highlight the potential benefits of transitioning to electric heating systems like heat pumps. Further policies derived from key state-level determinants promote attached housing, efficient setpoint temperature behaviors, energy-efficient appliances and lighting, and on-site electricity generation.