Recent efforts to reduce residential energy consumption have renewed interest in investigating salient drivers of household energy use. This article contributes to the ongoing literature by developing a model for examining geographic effects on energy use. Using a new, rich, micro-level survey that compiles information about dwelling attributes, occupant characteristics, and behaviors, we suggest a combined bottom-up and top-down statistical approach based on a multilevel regression model (MRM) and an innovative variable selection approach via the Adaptive Elastic Net Regularization technique (AdaEnetR). This approach enables us to extract geographic effects from the total variation in residential energy consumption and simultaneously explain the remaining variation with relevant explanatory variables and their interactions. The current model addresses several interrelated issues posed by the use of econometric methods to examine residential energy demand, including the risk of aggregation/disaggregation bias. Our empirical findings demonstrate the MRM's ability to effectively quantify approximately 0.67% of the geographic effects (aggregate level) and approximately 0.31% of the household and dwelling effects (individual level). Further, the findings show that household attributes are important factors that influence residential energy consumption patterns.