This study analyzes the efficiency of household-level energy consumption using a rich microdata set of homes within the United States. We measure efficiency by extending a cost-minimization model that treats the total amount of energy services produced as latent or unobserved due to technological differences in household consumption. The empirical strategy consists of applying latent class modeling to cost frontier analysis, which helps to identify heterogeneous subsets of units that require the fewest energy resources. Our estimates of efficient units form an empirical cost frontier of best practices within each subset. In order to understand the determinants of household-level energy efficiency, we condition the cost frontier analysis on numerous physical, climate-related, and socio-economic characteristics of the household. We find that state-level energy building code regulations, on average, induce a one-to-four percent marginal increase in household energy consumption.