Urban building energy modeling (UBEM) is a valuable tool for analyzing the building stock. Many different model approaches have been proposed in recent years suggesting various ways of dealing with the challenges of UBEM; however, central for all modeling approaches is the need for informative input data about the building stock. The availability of data for urban-scale modeling is both country-specific, time-consuming to aggregate, and data access is often limited due to privacy constraints. In this paper, we present a hierarchical bottom-up model of the Danish residential building stock using public building data for predicting the annual heating energy consumption. For more than 10,000 randomly selected single-family dwellings, the annual energy consumption is modeled and validated for the city of Aarhus, Denmark. We found that approx. 50% of the energy use is explained using only four widely available building characteristics, which enables building-scale predictions with a mean absolute error of approx. 25%. In addition, for city-scale predictions, the regression-based model enables aggregated predictions with a mean bias error of less than +/- 2%. Even though building-scale predictions are only somewhat accurate, the performance remains comparable to state-of-the art high-fidelity models in the literature. (C) 2018 Elsevier B.V. All rights reserved.