In a previous study, the collapse frequencies and associated mortalities of buildings were compiled based on various publications. The investigation showed a considerable scattering of collapse frequencies depending on the countries. In this paper, the collapse frequencies determined for buildings are compared with the results of probabilistic calculations. Such comparisons have already been carried out for bridges, dams, tunnels and retaining walls including the consideration of central estimators and standard deviations. In order to limit the scatter, the probabilities of failure for buildings were subdivided into the different causes of failure, and collapse frequencies were subdivided into different countries and geographical regions. Overall, the comparison shows that the probabilities of failure are on overage larger than the observed collapse frequencies. Furthermore, the comparison shows are large span, which is unusual for other types of structures. Of concern are the high observed collapse frequencies in various developing countries. Human error, lack of training, etc. are often cited as the cause. Should this correspond to the facts, the basic safety concept of modern building standards, which generally excludes human error, would only be applicable in these regions to a limited extent. However, the investigation includes some limitations, such as different safety targets in different standards and different years of constructions, which are not considered. Further work is required.