Climatic conditions are commonly considered the primary determinant of consumers' choices about energy use for heating. Besides, current regulations on the matter focus on the physical characteristics of the buildings, relying on a strict relationship between efficiency and savings. Nevertheless, the literature shows that energy demand determinants are difficult to be estimated with the accuracy required for predictive purposes, while the energy savings stemming from efficiency gains are partly outweighed by the consumers' behavior. We deal with these issues by analyzing spatial series data of natural gas consumptions for space heating and hot water production in the residential sector. The regression analysis takes four fields of covariates into account: climate, building characteristics, market aspects, and technological development. The estimation process is based on the following cornerstones: a spatially lagged dependent variable to deal with the problem of spatial autocorrelation, linear and logarithmic functional forms, and a two-stage interpolation strategy that is meant to provide unbiased estimates of both the dispersion matrix and the t-statistics by combining Ordinary Least Squares and Weighted Least Squares. The models turn out to be well specified, and their explanatory power is high, so the results are suitable for demand forecasting. Although the spatial autoregressive term does not appear among the significant regressors, we show that space does matter in shaping the natural gas consumption in different regions. Our analysis proves that heating gas demand is characterized by a positive elasticity to income. We use these results to provide estimates of the rebound effect. Nevertheless, the additional consumption directly attributable to an income effect is of moderate magnitude when the gas price does not vary. (C) 2017 Elsevier Ltd. All rights reserved.