Different views of corruption are discussed in the literature, ranging from theories highlighting a positive role of corruption (to 'grease the wheels' of an economy) to negative (acting as 'sand in the wheels') or even destructive effects of corruption on economic outcomes. The empirical evidence in this context is mixed, with alternative theories being supported by different studies, typically relying on different data sources, time periods, and measures of corruption. In this paper, we assess several alternative views of corruption simultaneously in a single empirical framework. Specifically, we test six prominent hypotheses regarding the role of corruption for government revenues using a comprehensive country-level panel dataset covering 194 countries in the period 1996-2019. Our results help to shed light on the factors that are empirically important in explaining the link between corruption and government revenues, including the role of governance, autocracy, fragile states, and natural resources.