The French term ‘cohabitation’ is commonly used to describe situations in semi-presidential systems where the prime minister and the president represent different political parties. The present contribution sets out to test to what extent cohabitation affects the powers of the president. Theoretically, the link between the phenomena in question is complex; there are arguments both for an assumption that cohabitation enhances the powers of the president and for a presumption that cohabitation is linked to decreasing powers of the head of state. It is far from self-evident how cohabitation should be operationalized, however. For instance, a president and a prime minister can represent different parties, but the parties can have a long tradition of working together in government coalitions. In some cases, the president and prime minister represent different parties, but the president’s party is included in the coalition government. Another variant is that either the president or the prime minister (or even both) are unaffiliated with political parties. In the framework of the present study, we make use of several operationalizations of cohabitation to assess to what extent the relationship between cohabitation and presidential powers is affected by whether cohabitation is broadly or narrowly defined. Empirically, the study is extensive in time and space. The research population consists of all democratic republics with a separate president and prime minister during the time period 1850-2021. In the literature, cohabitation is a term generally associated with semi-presidential forms of governments in which the president is popularly elected. The present paper takes a broader view and accordingly sets out to test if cohabitation is linked to presidential powers both in dual executives with popularly elected presidents and in systems with indirectly elected presidents. Since constitutional powers tend to remain unaltered irrespective of whether periods of cohabitation occur or not, the present study measures presidential powers with reference to actual, or ‘real’ powers. The powers of the president are measured with reference to seven questions in the V-dem dataset. Three of the questions refer explicitly to powers in the executive sphere, two to legislative powers, and two to powers that do not fall explicitly in either of the categories.