We employ a newly assembled indicator of corruption from Varieties of Democracy (V-Dem) to examine the effects of corruption on economic growth. The V-Dem indicator is coded for almost all contemporary and historical polities since the year 1900 and, for some countries, since the French Revolution. This extensive data source allows us to exploit long-run, slow-moving variation within countries for identification, circumventing many of the difficulties faced by previous studies based on cross-section data or short panels. We present robust evidence of a negative, linear causal effect of corruption on the steady-state rate of growth. Corruption, however, interacts with the institutional environment, giving rise to substantial effects heterogeneities in line with a weak form of the ‘grease the wheels’ hypothesis. In particular, corruption is found to be significantly more deleterious for growth in democracies than in autocracies. Our findings suggest that anti-corruption efforts should be targeted to the countries where corruption is more harmful for development.