We examine the effect of four cognitive biases related to occupants energy-behaviour on the energy performance gap by applying regression analysis to data collected from a pilot energy conservation program located in northeastern France. To avoid a set of potential problems in observational studies, we draw on predictive modelling tools, multiple imputation method, and inverse probability weighting strategy. Our results manifest the presence of a status quo bias referring to manual ventilation behaviour and an optimism bias associated with occupants' energy consumption behaviour yielding significantly higher net losing energy savings. Replacing missing observations with imputed values confirms these empirical findings and suggests evidence for an attitude-behaviour gap whereupon occupants with high pro-environmental attitudes do not significantly differ from others. Controlling for endogeneity reveals the presence of an intention-behaviour gap related to occupants' ecological motives and corroborates the robustness of the other findings. We conclude that self-reported energy behaviour and environmental attitudes might be a valid measurement for occupants' cognitive biases.