We used logistic regression to compare a set of habitat features inside known Scandinavian wolf Canis lupus territories with the same habitat features in areas outside known territories, but still close enough to be available for wolf colonization. In addition, we analysed changes in habitat variables over time within wolf territories. Wolf territories had lower densities of roads, built-up areas and open land than areas outside wolf territories, but there was no difference in the density of the wolves' main prey, elk Alces alces. The logistic regression model classified 79% of Scandinavia outside the reindeer husbandry area as suitable wolf habitat, that is with a probability of wolf occurrence >0.5. The proportion of built-up areas within the wolf territory decreased as the ‘borders’ of the wolf territory changed over time. Our model had a reasonably high predictive power, with correct classification in 90% (18 of 20) of the observed wolf territories in the study area. Polygons, randomly distributed outside the observed wolf territories, were correctly classified as not being occupied by wolves in 85% of the cases (17 of 20). This allows a more effective use of resources to, for example, prevent wolf depredation on livestock and dogs.