What determines whether autonomy ‘works’? Many questions around the alleged promises of vertical distribution of power are currently left unanswered. This article introduces a new global dataset on territorial self-governance arrangements currently covering 152 countries with 1426 regions for the period from 1950 to 2016. It contains more than 70,000 region-year assessments of the degree of autonomy of individual regions by applying the conceptual distinction between ‘self-rule’ and ‘shared-rule’ to all first-level administrative units globally. The central innovations are an explicit focus on both formal and informal territorial arrangements in democratic and non-democratic countries. A first glance at the main empirical patterns indicates that the whole variety of territorial self-governance arrangements has spread into all world regions – and increasingly so over the last two decades. Overall, there is considerably large variation regarding how much self-rule and especially shared-rule regions enjoy, both between and within different types of arrangements. Federal countries on average show the highest level of both self- and shared-rule indicators, followed by special autonomies with relatively high levels of self-rule and comparably lower levels of shared-rule. De facto autonomies, on the other hand, exhibit very high levels of self-rule combined with close to no shared-rule at all. With this data, it is now possible to account for subnational variation in autonomy and investigate the origins and effects of different institutional constellations, thus making it possible to address several important unanswered questions related to, for instance, self-determination conflicts, democracy and development.