We modelled presence/absence per tree of beetles and pseudoscorpions living in tree hollows in relation to trunk circumference, habitat openness, and connectivity (= density of hollow oaks in the surrounding area), using data from 281 oaks. The presence/absence models were then used to predict species’ occurrences in a county (11,600 km2) in southeastern Sweden. For eight of the nine species, the most parsimonious occupancy model included a positive relationship with connectivity and at least one tree characteristic. Occupancy underestimates from occurrence records—the ratio of the area of occupancy based on our predictive model to the area of occupancy based on occurrence records—varied between 3 and 83 among species when using occurrence records up to 1993, with significantly larger underestimates for smaller beetle species. Today (after extensive surveys), underestimation has decreased to 1.3–25, confirming that calculations solely based on species occurrence records greatly underestimate the area of occupancy. We suggest this should be taken into account to a greater extent and in a clearer way than today when constructing red lists. The radius of the connectivity measure that generated the best fit varied between 135 and 2,857 m among species, with longer distances for more threatened species. Consequently, preservation of the most threatened species (Elater ferrugineus and Tenebrio opacus) requires conservation efforts at larger spatial scales than required to protect Osmoderma eremita, which frequently has been used as an indicator and umbrella species.